question
stringlengths
101
4.47k
chosen
stringlengths
3
4.25k
rejected
stringlengths
39
9.29k
FIRST provide a one-sentence comparison of the two responses and explain \ which you feel is more helpful. SECOND, on a new line, state only "A" or "B" to indicate which response is more helpful.
๋จผ์ € ๋‘ ์‘๋‹ต์„ ํ•œ ๋ฌธ์žฅ์œผ๋กœ ๋น„๊ตํ•˜๊ณ , ์–ด๋–ค ์‘๋‹ต์ด ๋” ๋„์›€์ด ๋˜๋Š”์ง€ ์„ค๋ช…ํ•˜์„ธ์š”. ๋‘ ๋ฒˆ์งธ๋กœ, ์ƒˆ๋กœ์šด ์ค„์— "A" ๋˜๋Š” "B"๋งŒ ์ ์–ด ์–ด๋–ค ์‘๋‹ต์ด ๋” ๋„์›€์ด ๋˜๋Š”์ง€ ํ‘œ์‹œํ•˜์„ธ์š”.
๋จผ์ € ๋‘ ๋‹ต๋ณ€์„ ํ•œ ๋ฌธ์žฅ์œผ๋กœ ๋น„๊ตํ•˜๊ณ  ์–ด๋–ค ๋‹ต๋ณ€์ด ๋” ๋„์›€์ด ๋œ๋‹ค๊ณ  ์ƒ๊ฐํ•˜๋Š”์ง€ ์„ค๋ช…ํ•˜์„ธ์š”. ๋‘ ๋ฒˆ์งธ๋กœ, ์ƒˆ ์ค„์— "A" ๋˜๋Š” "B"๋ฅผ ํ‘œ์‹œํ•˜์—ฌ ์–ด๋–ค ๋‹ต๋ณ€์ด ๋” ๋„์›€์ด ๋˜๋Š”์ง€ ํ‘œ์‹œํ•ฉ๋‹ˆ๋‹ค.
Figure 3: The impact of mask ratio when training on Llama2-7B with GSM8K. We compare MFT of two settings: Fixed mask ratio without warm-up, and a linear warm-up of the mask ratio starting from 0.
Figure 3: GSM8K๋กœ Llama2-7B๋ฅผ ํ›ˆ๋ จํ•  ๋•Œ ๋งˆ์Šคํฌ ๋น„์œจ์˜ ์˜ํ–ฅ. ์šฐ๋ฆฌ๋Š” ๋‘ ๊ฐ€์ง€ ์„ค์ •์˜ MFT๋ฅผ ๋น„๊ตํ•ฉ๋‹ˆ๋‹ค: ์›Œ๋ฐ์—… ์—†์ด ๊ณ ์ •๋œ ๋งˆ์Šคํฌ ๋น„์œจ๊ณผ 0์—์„œ ์‹œ์ž‘ํ•˜๋Š” ๋งˆ์Šคํฌ ๋น„์œจ์˜ ์„ ํ˜• ์›Œ๋ฐ์—….
๊ทธ๋ฆผ 3: GSM8K๋กœ Llama2-7B๋ฅผ ํ›ˆ๋ จํ•  ๋•Œ ๋งˆ์Šคํฌ ๋น„์œจ์˜ ์˜ํ–ฅ. ๋‘ ์„ค์ •์˜ MFT๋ฅผ ๋น„๊ตํ•ฉ๋‹ˆ๋‹ค: ์›Œ๋ฐ์—…์ด ์—†๋Š” ๊ณ ์ • ๋งˆ์Šคํฌ ๋น„์œจ๊ณผ 0์—์„œ ์‹œ์ž‘ํ•˜๋Š” ๋งˆ์Šคํฌ ๋น„์œจ์˜ ์„ ํ˜• ์›Œ๋ฐ์—…์„ ๋น„๊ตํ•ฉ๋‹ˆ๋‹ค.
EvaluationWe consider two baselines in our experiments: LLaVA-1.6-Mistral-7B [31], one of our source models, and Japanese Stable VLM [43] a Japanese VLM trained from scratch on Japanese datasets.
ํ‰๊ฐ€ ์‹คํ—˜์—์„œ๋Š” ๋‘ ๊ฐ€์ง€ ๊ธฐ์ค€ ๋ชจ๋ธ์„ ๊ณ ๋ คํ•ฉ๋‹ˆ๋‹ค: ์†Œ์Šค ๋ชจ๋ธ ์ค‘ ํ•˜๋‚˜์ธ LLaVA-1.6-Mistral-7B [31]์™€ ์ผ๋ณธ์–ด ๋ฐ์ดํ„ฐ์…‹์œผ๋กœ ์ฒ˜์Œ๋ถ€ํ„ฐ ํ•™์Šต๋œ ์ผ๋ณธ์–ด VLM์ธ Japanese Stable VLM [43]์ž…๋‹ˆ๋‹ค.
ํ‰๊ฐ€์‹คํ—˜์—์„œ๋Š” ์†Œ์Šค ๋ชจ๋ธ ์ค‘ ํ•˜๋‚˜์ธ LLaVA-1.6-Mistral-7B [31]์™€ ์ผ๋ณธ ๋ฐ์ดํ„ฐ ์„ธํŠธ์—์„œ ์ฒ˜์Œ๋ถ€ํ„ฐ ํ•™์Šต๋œ ์ผ๋ณธ์–ด VLM์ธ ์ผ๋ณธ์–ด Stable VLM [43]์˜ ๋‘ ๊ฐ€์ง€ ๋ฒ ์ด์Šค๋ผ์ธ์„ ๊ณ ๋ คํ•ฉ๋‹ˆ๋‹ค.
Figure 7: Percentage of the most frequent entities for each fact composition type of TwoHopFact. The expanded forms of the abbreviations used for the fact composition types are listed in Table 4.
Figure 7: TwoHopFact์˜ ๊ฐ ์‚ฌ์‹ค ๊ตฌ์„ฑ ์œ ํ˜•์— ๋Œ€ํ•œ ๊ฐ€์žฅ ๋นˆ๋ฒˆํ•œ ์—”ํ‹ฐํ‹ฐ์˜ ๋น„์œจ. ์‚ฌ์‹ค ๊ตฌ์„ฑ ์œ ํ˜•์— ์‚ฌ์šฉ๋œ ์•ฝ์–ด์˜ ํ™•์žฅํ˜•์€ Table 4์— ๋‚˜์™€ ์žˆ์Šต๋‹ˆ๋‹ค.
๊ทธ๋ฆผ 7: TwoHopFact์˜ ๊ฐ ์‚ฌ์‹ค ๊ตฌ์„ฑ ์œ ํ˜•์— ๋Œ€ํ•œ ๊ฐ€์žฅ ๋นˆ๋ฒˆํ•œ ์—”ํ‹ฐํ‹ฐ์˜ ๋ฐฑ๋ถ„์œจ. ์‚ฌ์‹ค ๊ตฌ์„ฑ ์œ ํ˜•์— ์‚ฌ์šฉ๋œ ์•ฝ์–ด์˜ ํ™•์žฅ๋œ ํ˜•ํƒœ๋Š” ํ‘œ 4์— ๋‚˜์™€ ์žˆ์Šต๋‹ˆ๋‹ค.
In this experiment, our objective is to once again evaluate the effectiveness of the Adapt-LLM model, this time focusing on its ability to accurately determine when additional context is needed.
์ด๋ฒˆ ์‹คํ—˜์—์„œ ์šฐ๋ฆฌ์˜ ๋ชฉํ‘œ๋Š” ์ถ”๊ฐ€์ ์ธ ๋ฌธ๋งฅ์ด ํ•„์š”ํ•  ๋•Œ๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ํŒ๋‹จํ•˜๋Š” Adapt-LLM ๋ชจ๋ธ์˜ ํšจ๊ณผ๋ฅผ ๋‹ค์‹œ ํ•œ ๋ฒˆ ํ‰๊ฐ€ํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.
์ด๋ฒˆ ์‹คํ—˜์—์„œ๋Š” ์ถ”๊ฐ€ ์ปจํ…์ŠคํŠธ๊ฐ€ ํ•„์š”ํ•œ ์‹œ๊ธฐ๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ํŒ๋‹จํ•˜๋Š” ๋Šฅ๋ ฅ์— ์ดˆ์ ์„ ๋งž์ถฐ Adapt-LLM ๋ชจ๋ธ์˜ ํšจ๊ณผ๋ฅผ ๋‹ค์‹œ ํ•œ ๋ฒˆ ํ‰๊ฐ€ํ•˜๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•ฉ๋‹ˆ๋‹ค.
_Remark 3.4_ (why "sum of 3").: It is essential to obtain a lower bound that is the _sum_ of the three components; neglecting any may result in a suboptimal bound (see examples in Appendix A.4).
_์ฃผ์„ 3.4_ (์™œ "3์˜ ํ•ฉ"์ธ๊ฐ€).: ์„ธ ๊ฐ€์ง€ ๊ตฌ์„ฑ ์š”์†Œ์˜ _ํ•ฉ_์ธ ํ•˜ํ•œ์„ ์–ป๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค. ์–ด๋Š ํ•˜๋‚˜๋ผ๋„ ๋ฌด์‹œํ•˜๋ฉด ์ตœ์  ์ดํ•˜์˜ ๊ฒฝ๊ณ„๊ฐ€ ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค (๋ถ€๋ก A.4์˜ ์˜ˆ์‹œ๋ฅผ ์ฐธ์กฐํ•˜์‹ญ์‹œ์˜ค).
_Remark 3.4_ ("3์˜ ํ•ฉ"์ด์œ ): ์„ธ ๊ฐ€์ง€ ๊ตฌ์„ฑ ์š”์†Œ์˜ _ํ•ฉ๊ณ„_์ธ ํ•˜ํ•œ์„ ์–ป๋Š” ๊ฒƒ์ด ํ•„์ˆ˜์ ์ด๋ฉฐ, ์–ด๋Š ํ•˜๋‚˜๋ผ๋„ ๋ฌด์‹œํ•˜๋ฉด ์ตœ์ ์ด ์•„๋‹Œ ๋ฐ”์šด๋“œ๊ฐ€ ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค(๋ถ€๋ก A.4์˜ ์˜ˆ ์ฐธ์กฐ).
We hope future work can leverage CodecLM as a flexible data synthesis framework for LLM alignment, so that advances in the field can be integrated into CodecLM to reduce its current limitations.
์šฐ๋ฆฌ๋Š” ํ–ฅํ›„ ์—ฐ๊ตฌ๊ฐ€ CodecLM์„ LLM ์ •๋ ฌ์„ ์œ„ํ•œ ์œ ์—ฐํ•œ ๋ฐ์ดํ„ฐ ํ•ฉ์„ฑ ํ”„๋ ˆ์ž„์›Œํฌ๋กœ ํ™œ์šฉํ•˜์—ฌ, ์ด ๋ถ„์•ผ์˜ ๋ฐœ์ „์ด CodecLM์— ํ†ตํ•ฉ๋˜์–ด ํ˜„์žฌ์˜ ํ•œ๊ณ„๋ฅผ ์ค„์ผ ์ˆ˜ ์žˆ๊ธฐ๋ฅผ ๋ฐ”๋ž๋‹ˆ๋‹ค.
ํ–ฅํ›„ ์—ฐ๊ตฌ์—์„œ CodecLM์„ LLM ์ •๋ ฌ์„ ์œ„ํ•œ ์œ ์—ฐํ•œ ๋ฐ์ดํ„ฐ ํ•ฉ์„ฑ ํ”„๋ ˆ์ž„์›Œํฌ๋กœ ํ™œ์šฉํ•˜์—ฌ ์ด ๋ถ„์•ผ์˜ ๋ฐœ์ „์„ CodecLM์— ํ†ตํ•ฉํ•˜์—ฌ ํ˜„์žฌ์˜ ํ•œ๊ณ„๋ฅผ ์ค„์ผ ์ˆ˜ ์žˆ๊ธฐ๋ฅผ ๊ธฐ๋Œ€ํ•œ๋‹ค.
Here, \(W_{K}\in\mathbb{R}^{d_{model}\times d_{bay}}\), \(W_{V}\in\mathbb{R}^{d_{model}\times d_{value}}\) and \(W_{Q}\in\mathbb{R}^{d_{model}\times d_{bay}}\) are trainable projection matrices.
์—ฌ๊ธฐ์„œ \(W_{K}\in\mathbb{R}^{d_{model}\times d_{bay}}\), \(W_{V}\in\mathbb{R}^{d_{model}\times d_{value}}\) ๋ฐ \(W_{Q}\in\mathbb{R}^{d_{model}\times d_{bay}}\)๋Š” ํ•™์Šต ๊ฐ€๋Šฅํ•œ ํˆฌ์˜ ํ–‰๋ ฌ์ž…๋‹ˆ๋‹ค.
์—ฌ๊ธฐ์„œ \(W_{K}\in\mathbb{R}^{d_{model}\times d_{bay}}\), \(W_{V}\in\mathbb{R}^{d_{model}\times d_{value}}\), \(W_{Q}\in\mathbb{R}^{d_{model}\times d_{bay}}\)๋Š” ํ•™์Šต ๊ฐ€๋Šฅํ•œ ํˆฌ์˜ ํ–‰๋ ฌ์ž…๋‹ˆ๋‹ค.
Figure 3: Average compute (GFLOPS) required to process a single word within (a) multilingual, (b) English, and (c) German documents within a full **training** pass (including the backward pass).
Figure 3: ์ „์ฒด **ํ›ˆ๋ จ** ๊ณผ์ •(์—ญ์ „ํŒŒ ํฌํ•จ)์—์„œ (a) ๋‹ค๊ตญ์–ด, (b) ์˜์–ด, (c) ๋…์ผ์–ด ๋ฌธ์„œ ๋‚ด ๋‹จ์–ด ํ•˜๋‚˜๋ฅผ ์ฒ˜๋ฆฌํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ ํ‰๊ท  ์—ฐ์‚ฐ๋Ÿ‰(GFLOPS).
๊ทธ๋ฆผ 3: ์ „์ฒด **ํ›ˆ๋ จ** ํŒจ์Šค(์—ญ๋ฐฉํ–ฅ ํŒจ์Šค ํฌํ•จ) ๋‚ด์—์„œ (a) ๋‹ค๊ตญ์–ด, (b) ์˜์–ด ๋ฐ (c) ๋…์ผ์–ด ๋ฌธ์„œ ๋‚ด์—์„œ ๋‹จ์ผ ๋‹จ์–ด๋ฅผ ์ฒ˜๋ฆฌํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ ํ‰๊ท  ๊ณ„์‚ฐ(GFLOPS)์ž…๋‹ˆ๋‹ค.
Release requirementsTo ensure the integrity of test results, model publishers (i.e., organizations who make SUTs available) commit to adhering to the following rules, which may change over time:
์ถœ์‹œ ์š”๊ตฌ ์‚ฌํ•ญ: ํ…Œ์ŠคํŠธ ๊ฒฐ๊ณผ์˜ ๋ฌด๊ฒฐ์„ฑ์„ ๋ณด์žฅํ•˜๊ธฐ ์œ„ํ•ด ๋ชจ๋ธ ๊ฒŒ์‹œ์ž(์ฆ‰, SUT๋ฅผ ์ œ๊ณตํ•˜๋Š” ์กฐ์ง)๋Š” ๋‹ค์Œ ๊ทœ์น™์„ ์ค€์ˆ˜ํ•  ๊ฒƒ์„ ์•ฝ์†ํ•˜๋ฉฐ, ์ด ๊ทœ์น™์€ ์‹œ๊ฐ„์ด ์ง€๋‚จ์— ๋”ฐ๋ผ ๋ณ€๊ฒฝ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค:
๋ฆด๋ฆฌ์Šค ์š”๊ตฌ ์‚ฌํ•ญํ…Œ์ŠคํŠธ ๊ฒฐ๊ณผ์˜ ๋ฌด๊ฒฐ์„ฑ์„ ๋ณด์žฅํ•˜๊ธฐ ์œ„ํ•ด ๋ชจ๋ธ ๊ฒŒ์‹œ์ž(์˜ˆ: SUT์„ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•˜๋Š” ์กฐ์ง)๋Š” ์‹œ๊ฐ„์ด ์ง€๋‚จ์— ๋”ฐ๋ผ ๋ณ€๊ฒฝ๋  ์ˆ˜ ์žˆ๋Š” ๋‹ค์Œ ๊ทœ์น™์„ ์ค€์ˆ˜ํ•˜๊ธฐ๋กœ ์•ฝ์ •ํ•ฉ๋‹ˆ๋‹ค:
Footnote 20: The modelโ€™s ability to learn from junk data is negligible; each person in bioS(\(N^{\prime}\)) appears only 0.2 times during training when \(N=200k\), or 0.05 times when \(N=50k\).
Footnote 20: ๋ชจ๋ธ์ด ๋ถˆํ•„์š”ํ•œ ๋ฐ์ดํ„ฐ๋กœ๋ถ€ํ„ฐ ํ•™์Šตํ•˜๋Š” ๋Šฅ๋ ฅ์€ ๋ฌด์‹œํ•  ๋งŒํ•ฉ๋‹ˆ๋‹ค. bioS(\(N^{\prime}\))์˜ ๊ฐ ์‚ฌ๋žŒ์€ \(N=200k\)์ผ ๋•Œ ํ›ˆ๋ จ ์ค‘์— ๋‹จ 0.2๋ฒˆ, \(N=50k\)์ผ ๋•Œ๋Š” 0.05๋ฒˆ๋งŒ ๋‚˜ํƒ€๋‚ฉ๋‹ˆ๋‹ค.
๊ฐ์ฃผ 20: ๋ชจ๋ธ์ด ์ •ํฌ ๋ฐ์ดํ„ฐ์—์„œ ํ•™์Šตํ•˜๋Š” ๋Šฅ๋ ฅ์€ ๋ฏธ๋ฏธํ•œ ์ˆ˜์ค€์ด๋ฉฐ, bioS(\(N^{\ํ”„๋ผ์ž„}\))์˜ ๊ฐ ์‚ฌ๋žŒ์€ \(N=200k\)์ผ ๋•Œ ํ›ˆ๋ จ ์ค‘ 0.2ํšŒ, \(N=50k\)์ผ ๋•Œ 0.05ํšŒ๋งŒ ๋‚˜ํƒ€๋‚ฉ๋‹ˆ๋‹ค.
Generation TaskDeactivation methods: (i) randomly sampled neurons in the generating layers. (ii) randomly sampled neurons in all layers. (iv) language-specific neurons in the generating layers.
์ƒ์„ฑ ์ž‘์—… ๋น„ํ™œ์„ฑํ™” ๋ฐฉ๋ฒ•: (i) ์ƒ์„ฑ ๋ ˆ์ด์–ด์—์„œ ๋ฌด์ž‘์œ„๋กœ ์ƒ˜ํ”Œ๋ง๋œ ๋‰ด๋Ÿฐ. (ii) ๋ชจ๋“  ๋ ˆ์ด์–ด์—์„œ ๋ฌด์ž‘์œ„๋กœ ์ƒ˜ํ”Œ๋ง๋œ ๋‰ด๋Ÿฐ. (iv) ์ƒ์„ฑ ๋ ˆ์ด์–ด์—์„œ ์–ธ์–ด๋ณ„ ๋‰ด๋Ÿฐ.
์„ธ๋Œ€ ์ž‘์—…๋น„ํ™œ์„ฑํ™” ๋ฐฉ๋ฒ•: (i) ์ƒ์„ฑ ๋ ˆ์ด์–ด์—์„œ ๋ฌด์ž‘์œ„๋กœ ์ƒ˜ํ”Œ๋ง๋œ ๋‰ด๋Ÿฐ. (ii) ๋ชจ๋“  ๋ ˆ์ด์–ด์—์„œ ๋ฌด์ž‘์œ„๋กœ ์ƒ˜ํ”Œ๋ง๋œ ๋‰ด๋Ÿฐ. (iv) ์ƒ์„ฑ ๋ ˆ์ด์–ด์˜ ์–ธ์–ด๋ณ„ ๋‰ด๋Ÿฐ.
Figure 9: \(|\) R-GSM example where the original problem can be correctly solved by all LLMs, but GPT-3.5-Turbo fails on the reordered version while all the other LLMs still solve it correctly.
Figure 9: ์›๋ž˜ ๋ฌธ์ œ๋Š” ๋ชจ๋“  LLM์ด ์˜ฌ๋ฐ”๋ฅด๊ฒŒ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, ์žฌ์ •๋ ฌ๋œ ๋ฒ„์ „์—์„œ๋Š” GPT-3.5-Turbo๊ฐ€ ์‹คํŒจํ•˜๊ณ  ๋‹ค๋ฅธ ๋ชจ๋“  LLM์€ ์—ฌ์ „ํžˆ ์˜ฌ๋ฐ”๋ฅด๊ฒŒ ํ•ด๊ฒฐํ•˜๋Š” R-GSM ์˜ˆ์ œ์ž…๋‹ˆ๋‹ค.
๊ทธ๋ฆผ 9: ์›๋ž˜ ๋ฌธ์ œ๋Š” ๋ชจ๋“  LLM์ด ์˜ฌ๋ฐ”๋ฅด๊ฒŒ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, ์ˆœ์„œ๊ฐ€ ๋ฐ”๋€ ๋ฒ„์ „์—์„œ๋Š” GPT-3.5-Turbo๊ฐ€ ์‹คํŒจํ•˜๋Š” ๋ฐ˜๋ฉด ๋‹ค๋ฅธ ๋ชจ๋“  LLM์€ ์—ฌ์ „ํžˆ ์˜ฌ๋ฐ”๋ฅด๊ฒŒ ํ•ด๊ฒฐํ•˜๋Š” \(|\) R-GSM ์˜ˆ์ œ.
**Reading comprehension:** Reading a given article or paragraph and answering questions about it. We use RACE [23] for English and C3 [42] for Chinese. Both are based on multi-choice questions.
**๋…ํ•ด:** ์ฃผ์–ด์ง„ ๊ธฐ์‚ฌ๋‚˜ ๋‹จ๋ฝ์„ ์ฝ๊ณ  ์ด์— ๋Œ€ํ•œ ์งˆ๋ฌธ์— ๋‹ตํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์˜์–ด๋ฅผ ์œ„ํ•ด RACE [23]๋ฅผ, ์ค‘๊ตญ์–ด๋ฅผ ์œ„ํ•ด C3 [42]๋ฅผ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค. ๋‘˜ ๋‹ค ๋‹ค์ง€์„ ๋‹คํ˜• ์งˆ๋ฌธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค.
**๋…ํ•ด๋ ฅ: ์ฃผ์–ด์ง„ ๊ธฐ์‚ฌ๋‚˜ ๋‹จ๋ฝ์„ ์ฝ๊ณ  ๊ทธ์— ๋Œ€ํ•œ ์งˆ๋ฌธ์— ๋‹ตํ•˜๋Š” ๋ฌธ์ œ์ž…๋‹ˆ๋‹ค. ์˜์–ด๋Š” RACE [23], ์ค‘๊ตญ์–ด๋Š” C3 [42]๋ฅผ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค. ๋‘˜ ๋‹ค ๊ฐ๊ด€์‹ ๋ฌธ์ œ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค.
**Evaluation of MLLM.** After introducing the approaches to developing MLLMs, we further discuss how to effectively assess the multimodal capabilities of MLLMs from the following three aspects.
MLLM์˜ ํ‰๊ฐ€. MLLM์„ ๊ฐœ๋ฐœํ•˜๋Š” ์ ‘๊ทผ ๋ฐฉ์‹์„ ์†Œ๊ฐœํ•œ ํ›„, ๋‹ค์Œ ์„ธ ๊ฐ€์ง€ ์ธก๋ฉด์—์„œ MLLM์˜ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๊ธฐ๋Šฅ์„ ํšจ๊ณผ์ ์œผ๋กœ ํ‰๊ฐ€ํ•˜๋Š” ๋ฐฉ๋ฒ•์— ๋Œ€ํ•ด ๋…ผ์˜ํ•ฉ๋‹ˆ๋‹ค.
**MLLM์˜ ํ‰๊ฐ€**. MLLM ๊ฐœ๋ฐœ ์ ‘๊ทผ ๋ฐฉ์‹์„ ์†Œ๊ฐœํ•œ ํ›„, ๋‹ค์Œ ์„ธ ๊ฐ€์ง€ ์ธก๋ฉด์—์„œ MLLM์˜ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๊ธฐ๋Šฅ์„ ํšจ๊ณผ์ ์œผ๋กœ ํ‰๊ฐ€ํ•˜๋Š” ๋ฐฉ๋ฒ•์— ๋Œ€ํ•ด ์ž์„ธํžˆ ๋…ผ์˜ํ•ฉ๋‹ˆ๋‹ค.
Table 2 provides insights into the characteristics of the three datasets involved in our experimental procedure, including the total number of questions and the average number of words per ques
Table 2๋Š” ์‹คํ—˜ ์ ˆ์ฐจ์— ํฌํ•จ๋œ ์„ธ ๊ฐ€์ง€ ๋ฐ์ดํ„ฐ์…‹์˜ ํŠน์„ฑ์— ๋Œ€ํ•œ ํ†ต์ฐฐ์„ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์ด ์งˆ๋ฌธ ์ˆ˜์™€ ์งˆ๋ฌธ๋‹น ํ‰๊ท  ๋‹จ์–ด ์ˆ˜๊ฐ€ ํฌํ•จ๋ฉ๋‹ˆ๋‹ค.
ํ‘œ 2๋Š” ์ด ์งˆ๋ฌธ ์ˆ˜์™€ ์งˆ๋ฌธ๋‹น ํ‰๊ท  ๋‹จ์–ด ์ˆ˜๋ฅผ ํฌํ•จํ•˜์—ฌ ์‹คํ—˜ ์ ˆ์ฐจ์— ํฌํ•จ๋œ ์„ธ ๊ฐ€์ง€ ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ํŠน์„ฑ์— ๋Œ€ํ•œ ์ธ์‚ฌ์ดํŠธ๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.
2020; Wu et al., 2021; Xiao et al., 2021; Li et al., 2022; Gupta and Gupta, 2022) is more challenging as it introduces additional temporal information compared to visual multi-modal reasoning.
2020; Wu et al., 2021; Xiao et al., 2021; Li et al., 2022; Gupta and Gupta, 2022)๋Š” ์‹œ๊ฐ์  ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ์ถ”๋ก ์— ๋น„ํ•ด ์ถ”๊ฐ€์ ์ธ ์‹œ๊ฐ„ ์ •๋ณด๋ฅผ ๋„์ž…ํ•˜๊ธฐ ๋•Œ๋ฌธ์— ๋” ๋„์ „์ ์ž…๋‹ˆ๋‹ค.
์‹œ๊ฐ์  ๋‹ค์ค‘ ๋ชจ๋‹ฌ ์ถ”๋ก ์— ๋น„ํ•ด ์ถ”๊ฐ€์ ์ธ ์‹œ๊ฐ„์  ์ •๋ณด๋ฅผ ๋„์ž…ํ•˜๊ธฐ ๋•Œ๋ฌธ์— 2020; Wu et al., 2021; Xiao et al., 2021; Li et al., 2022; Gupta and Gupta, 2022)๋Š” ๋” ์–ด๋ ต๋‹ค.
In this part, we present discussions regarding two fundamental questions related to CoT prompting, _i.e._, _"when does CoT prompting work for LLMs"_ and _"why can LLMs perform CoT reasoning"_.
์ด ๋ถ€๋ถ„์—์„œ๋Š” CoT ํ”„๋กฌํ”„ํŠธ์™€ ๊ด€๋ จ๋œ ๋‘ ๊ฐ€์ง€ ๊ทผ๋ณธ์ ์ธ ์งˆ๋ฌธ, ์ฆ‰ "LLMs์—์„œ CoT ํ”„๋กฌํ”„ํŠธ๊ฐ€ ์–ธ์ œ ์ž‘๋™ํ•˜๋Š”๊ฐ€"์™€ "์™œ LLMs๊ฐ€ CoT ์ถ”๋ก ์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋Š”๊ฐ€"์— ๋Œ€ํ•œ ๋…ผ์˜๋ฅผ ์ œ์‹œํ•ฉ๋‹ˆ๋‹ค.
์ด ๋ถ€๋ถ„์—์„œ๋Š” CoT ํ”„๋กฌํ”„ํŠธ์™€ ๊ด€๋ จ๋œ ๋‘ ๊ฐ€์ง€ ๊ทผ๋ณธ์ ์ธ ์งˆ๋ฌธ, ์ฆ‰ _"์–ธ์ œ CoT ํ”„๋กฌํ”„ํŠธ๊ฐ€ LLM์— ์ž‘๋™ํ•˜๋Š”๊ฐ€"_์™€ _"์™œ LLM์ด CoT ์ถ”๋ก ์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋Š”๊ฐ€"_์— ๋Œ€ํ•œ ๋…ผ์˜๋ฅผ ์ œ์‹œํ•œ๋‹ค.
**S&P**Plausible but only partially supported. The Reference [1] does not explicitly mention the person is born in Moscow and there is no description about the Russian Guild of Film Producers.
S&P: ๊ทธ๋Ÿด๋“ฏํ•˜์ง€๋งŒ ๋ถ€๋ถ„์ ์œผ๋กœ๋งŒ ์ง€์›๋ฉ๋‹ˆ๋‹ค. ์ฐธ๊ณ  ๋ฌธํ—Œ [1]์—๋Š” ๊ทธ ์‚ฌ๋žŒ์ด ๋ชจ์Šคํฌ๋ฐ”์—์„œ ํƒœ์–ด๋‚ฌ๋‹ค๋Š” ๋ช…์‹œ์ ์ธ ์–ธ๊ธ‰์ด ์—†์œผ๋ฉฐ, ๋Ÿฌ์‹œ์•„ ์˜ํ™” ์ œ์ž‘์ž ๊ธธ๋“œ์— ๋Œ€ํ•œ ์„ค๋ช…๋„ ์—†์Šต๋‹ˆ๋‹ค.
**S&P**๊ทธ๋Ÿด๋“ฏํ•˜์ง€๋งŒ ๋ถ€๋ถ„์ ์œผ๋กœ๋งŒ ๋’ท๋ฐ›์นจ๋ฉ๋‹ˆ๋‹ค. ์ฐธ์กฐ [1]์—๋Š” ํ•ด๋‹น ์ธ๋ฌผ์ด ๋ชจ์Šคํฌ๋ฐ”์—์„œ ํƒœ์–ด๋‚ฌ๋‹ค๋Š” ์–ธ๊ธ‰์ด ๋ช…์‹œ์ ์œผ๋กœ ์—†์œผ๋ฉฐ ๋Ÿฌ์‹œ์•„ ์˜ํ™” ์ œ์ž‘์ž ๊ธธ๋“œ์— ๋Œ€ํ•œ ์„ค๋ช…์ด ์—†์Šต๋‹ˆ๋‹ค.
Our web documents in the corpora consist of Oscars1Abadji et al. (2021), that were generated by the ungoliant pipeline2 based on three Common Crawl WET Archives (2022-27, 2022-49 and 2023-14).
์šฐ๋ฆฌ ๋ง๋ญ‰์น˜์˜ ์›น ๋ฌธ์„œ๋Š” Oscars1Abadji et al. (2021)๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ์œผ๋ฉฐ, ์ด๋Š” ์„ธ ๊ฐœ์˜ Common Crawl WET Archives (2022-27, 2022-49, 2023-14)๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ungoliant ํŒŒ์ดํ”„๋ผ์ธ2์— ์˜ํ•ด ์ƒ์„ฑ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.
๋ง๋ญ‰์น˜์˜ ์›น ๋ฌธ์„œ๋Š” ์„ธ ๊ฐœ์˜ Common Crawl WET ์•„์นด์ด๋ธŒ(2022-27, 2022-49 ๋ฐ 2023-14)๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ungoliant ํŒŒ์ดํ”„๋ผ์ธ2์— ์˜ํ•ด ์ƒ์„ฑ๋œ Oscars1Abadji ๋“ฑ(2021)์œผ๋กœ ๊ตฌ์„ฑ๋ฉ๋‹ˆ๋‹ค.
Equal contribution. See author contributions for details. Work done during their internships at Microsoft Research Asia. EUR [email protected]; [email protected] authors.
๋™์ผํ•œ ๊ธฐ์—ฌ. ์ž์„ธํ•œ ๋‚ด์šฉ์€ ์ €์ž ๊ธฐ์—ฌ๋„๋ฅผ ์ฐธ์กฐํ•˜์‹ญ์‹œ์˜ค. Microsoft Research Asia์—์„œ ์ธํ„ด์‹ญ ๋™์•ˆ ์ˆ˜ํ–‰๋œ ์ž‘์—…์ž…๋‹ˆ๋‹ค. EUR [email protected]; [email protected] ์—ฐ๋ฝ ์ €์ž.
๋™๋“ฑํ•œ ๊ธฐ์—ฌ. ์ž์„ธํ•œ ๋‚ด์šฉ์€ ์ €์ž ๊ธฐ์—ฌ๋„๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”. Microsoft Research Asia์—์„œ ์ธํ„ด์œผ๋กœ ๊ทผ๋ฌดํ•œ ๊ธฐ๊ฐ„ ๋™์•ˆ ์ˆ˜ํ–‰ํ•œ ์ž‘์—…. EUR [email protected]; [email protected]๊ต์‹  ์ €์ž.
After the death of Jurning mother, lost and Mrs. Price grow slower and decides to nearly adults inside early experiences about lost and Mrs. Price wedding but later becomes very at this point.
Jurning ์–ด๋จธ๋‹ˆ์˜ ์ฃฝ์Œ ์ดํ›„, ์žƒ์–ด๋ฒ„๋ฆฐ ๊ฒƒ๊ณผ ํ”„๋ผ์ด์Šค ๋ถ€์ธ์€ ์ ์  ๋Š๋ ค์ง€๊ณ  ๊ฑฐ์˜ ์„ฑ์ธ์ด ๋˜๊ธฐ ์ „์˜ ์ดˆ๊ธฐ ๊ฒฝํ—˜์— ๋Œ€ํ•ด ๊ฒฐํ˜ผ์„ ์ƒ๊ฐํ•˜์ง€๋งŒ ๋‚˜์ค‘์—๋Š” ๋งค์šฐ ๋‹ฌ๋ผ์ง‘๋‹ˆ๋‹ค.
์ฃผ๋‹ ์–ด๋จธ๋‹ˆ๊ฐ€ ์ฃฝ์€ ํ›„, ์žƒ์–ด๋ฒ„๋ฆฐ ๊ฐ€๊ฒฉ ๋ถ€์ธ์€ ๋”๋””๊ฒŒ ์ž๋ผ๊ธฐ ์‹œ์ž‘ํ•˜๊ณ  ์žƒ์–ด๋ฒ„๋ฆฐ ๊ฐ€๊ฒฉ ๋ถ€์ธ์˜ ๊ฒฐํ˜ผ์‹์— ๋Œ€ํ•œ ์ดˆ๊ธฐ ๊ฒฝํ—˜์„ ํ†ตํ•ด ๊ฑฐ์˜ ์„ฑ์ธ์ด ๋˜์ง€๋งŒ ๋‚˜์ค‘์— ์ด ์‹œ์ ์—์„œ ๋งค์šฐ ๋ฉ๋‹ˆ๋‹ค.
The WMT competition winners for each direction as reported in WMT'21 and WMT'22 correspond to those used by Hendy et al. (2023). For more detailed information, we direct readers to this paper.
WMT'21 ๋ฐ WMT'22์—์„œ ๋ณด๊ณ ๋œ ๊ฐ ๋ฐฉํ–ฅ์˜ WMT ๋Œ€ํšŒ ์šฐ์Šน์ž๋Š” Hendy et al. (2023)์—์„œ ์‚ฌ์šฉ๋œ ๊ฒƒ๊ณผ ์ผ์น˜ํ•ฉ๋‹ˆ๋‹ค. ์ž์„ธํ•œ ์ •๋ณด๋Š” ์ด ๋…ผ๋ฌธ์„ ์ฐธ์กฐํ•˜์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.
WMT'21 ๋ฐ WMT'22์— ๋ณด๊ณ ๋œ ๊ฐ ๋ฐฉํ–ฅ์˜ WMT ๊ฒฝ์Ÿ ์šฐ์Šน์ž๋Š” Hendy ๋“ฑ(2023)์—์„œ ์‚ฌ์šฉํ•œ ๊ฒƒ๊ณผ ์ผ์น˜ํ•œ๋‹ค. ์ž์„ธํ•œ ๋‚ด์šฉ์€ ์ด ๋…ผ๋ฌธ์„ ์ฐธ์กฐํ•˜์‹ญ์‹œ์˜ค.
_(B) **Later-Stage Dynamics.** As soon as_ \(T\gtrsim\overline{T}\)_, it holds that_ \(E_{test}\asymp(\pi T)^{-c}\)_. Similarly, we recover the unpolluted sample-size law scaling_ \(T^{-c}\)_.
_(B) **ํ›„๊ธฐ ๋‹จ๊ณ„ ๋™์—ญํ•™.** \(T\gtrsim\overline{T}\)๊ฐ€ ๋˜๋ฉด, \(E_{test}\asymp(\pi T)^{-c}\)๊ฐ€ ์„ฑ๋ฆฝํ•ฉ๋‹ˆ๋‹ค. ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ, ์˜ค์—ผ๋˜์ง€ ์•Š์€ ์ƒ˜ํ”Œ ํฌ๊ธฐ ๋ฒ•์น™ \(T^{-c}\)๋ฅผ ํšŒ๋ณตํ•ฉ๋‹ˆ๋‹ค.
_(B) **ํ›„๊ธฐ ๋™์—ญํ•™**._ \(T\gtrsim\overline{T}\)_, it holds that_ \(E_{test}\asymp(\pi T)^{-c}\)_. ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ, ์˜ค์—ผ๋˜์ง€ ์•Š์€ ์ƒ˜ํ”Œ ํฌ๊ธฐ ๋ฒ•์น™ ์Šค์ผ€์ผ๋ง_ \(T^{-c}\)_.
Commonsense reasoning is one of the important capabilities each model can obtain. This capability denotes the ability of the model to use prior knowledge in combination with reasoning skills.
์ƒ์‹์  ์ถ”๋ก ์€ ๊ฐ ๋ชจ๋ธ์ด ์–ป์„ ์ˆ˜ ์žˆ๋Š” ์ค‘์š”ํ•œ ๋Šฅ๋ ฅ ์ค‘ ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค. ์ด ๋Šฅ๋ ฅ์€ ๋ชจ๋ธ์ด ์‚ฌ์ „ ์ง€์‹์„ ์ถ”๋ก  ๊ธฐ์ˆ ๊ณผ ๊ฒฐํ•ฉํ•˜์—ฌ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ๋Šฅ๋ ฅ์„ ๋‚˜ํƒ€๋ƒ…๋‹ˆ๋‹ค.
์ƒ์‹ ์ถ”๋ก ์€ ๊ฐ ๋ชจ๋ธ์ด ํš๋“ํ•  ์ˆ˜ ์žˆ๋Š” ์ค‘์š”ํ•œ ๊ธฐ๋Šฅ ์ค‘ ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค. ์ด ๊ธฐ๋Šฅ์€ ๋ชจ๋ธ์ด ์ถ”๋ก  ๊ธฐ์ˆ ๊ณผ ๊ฒฐํ•ฉํ•˜์—ฌ ์‚ฌ์ „ ์ง€์‹์„ ์‚ฌ์šฉํ•˜๋Š” ๋Šฅ๋ ฅ์„ ๋‚˜ํƒ€๋ƒ…๋‹ˆ๋‹ค.
By following these steps, it is possible to train an LLM using RLHF methods based on InstructGPT to create a human assistant that exceeds open assistant in performance, accuracy, and utility.
์ด๋Ÿฌํ•œ ๋‹จ๊ณ„๋ฅผ ๋”ฐ๋ฅด๋ฉด InstructGPT๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ RLHF ๋ฐฉ๋ฒ•์„ ์‚ฌ์šฉํ•˜์—ฌ ์„ฑ๋Šฅ, ์ •ํ™•์„ฑ ๋ฐ ์œ ์šฉ์„ฑ์—์„œ open assistant๋ฅผ ๋Šฅ๊ฐ€ํ•˜๋Š” ์ธ๊ฐ„ ๋ณด์กฐ์ž๋ฅผ ๋งŒ๋“ค๊ธฐ ์œ„ํ•ด LLM์„ ํ›ˆ๋ จ์‹œํ‚ฌ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
์ด ๋‹จ๊ณ„๋ฅผ ์ˆ˜ํ–‰ํ•˜๋ฉด InstructGPT๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ RLHF ๋ฐฉ๋ฒ•์„ ์‚ฌ์šฉํ•˜์—ฌ LLM์„ ํ›ˆ๋ จํ•˜์—ฌ ์„ฑ๋Šฅ, ์ •ํ™•์„ฑ ๋ฐ ์œ ์šฉ์„ฑ ์ธก๋ฉด์—์„œ ์˜คํ”ˆ ์–ด์‹œ์Šคํ„ดํŠธ๋ฅผ ๋Šฅ๊ฐ€ํ•˜๋Š” ์ธ๊ฐ„ ์–ด์‹œ์Šคํ„ดํŠธ๋ฅผ ๋งŒ๋“ค ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
**Will the dataset be distributed under a copyright or other intellectual property (IP) license and/or under applicable terms of use (ToU)?** The dataset is distributed under a CC-BY license.
**๋ฐ์ดํ„ฐ์…‹์ด ์ €์ž‘๊ถŒ ๋˜๋Š” ๊ธฐํƒ€ ์ง€์  ์žฌ์‚ฐ๊ถŒ (IP) ๋ผ์ด์„ ์Šค ๋ฐ/๋˜๋Š” ์ ์šฉ ๊ฐ€๋Šฅํ•œ ์ด์šฉ ์•ฝ๊ด€ (ToU) ํ•˜์— ๋ฐฐํฌ๋ฉ๋‹ˆ๊นŒ?** ๋ฐ์ดํ„ฐ์…‹์€ CC-BY ๋ผ์ด์„ ์Šค ํ•˜์— ๋ฐฐํฌ๋ฉ๋‹ˆ๋‹ค.
**๋ฐ์ดํ„ฐ ์„ธํŠธ๊ฐ€ ์ €์ž‘๊ถŒ ๋˜๋Š” ๊ธฐํƒ€ ์ง€์  ์žฌ์‚ฐ๊ถŒ(IP) ๋ผ์ด์„ ์Šค ๋ฐ/๋˜๋Š” ํ•ด๋‹น ์‚ฌ์šฉ ์•ฝ๊ด€(ToU)์— ๋”ฐ๋ผ ๋ฐฐํฌ๋˜๋‚˜์š”?** ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” CC-BY ๋ผ์ด์„ ์Šค์— ๋”ฐ๋ผ ๋ฐฐํฌ๋ฉ๋‹ˆ๋‹ค.
In sDPO, we propose to use the available preference datasets in a stepwise manner instead of using them all at once. The comparison of the overall flow of DPO and sDPO is presented in Fig. 1.
sDPO์—์„œ๋Š” ๋ชจ๋“  ์„ ํ˜ธ ๋ฐ์ดํ„ฐ์…‹์„ ํ•œ ๋ฒˆ์— ์‚ฌ์šฉํ•˜๋Š” ๋Œ€์‹  ๋‹จ๊ณ„์ ์œผ๋กœ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค. DPO์™€ sDPO์˜ ์ „์ฒด ํ๋ฆ„ ๋น„๊ต๋Š” Fig. 1์— ์ œ์‹œ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
sDPO์—์„œ๋Š” ์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ์„ ํ˜ธ๋„ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ํ•œ ๋ฒˆ์— ๋ชจ๋‘ ์‚ฌ์šฉํ•˜๋Š” ๋Œ€์‹  ๋‹จ๊ณ„๋ณ„ ๋ฐฉ์‹์œผ๋กœ ์‚ฌ์šฉํ•  ๊ฒƒ์„ ์ œ์•ˆํ•œ๋‹ค. DPO์™€ sDPO์˜ ์ „์ฒด ํ๋ฆ„ ๋น„๊ต๋Š” Fig. 1์— ๋‚˜์™€ ์žˆ๋‹ค.
A single head in the vanilla MHA computes its attention context \(A_{dot}\in\mathbb{R}^{N\times d_{value}}\) from sequence of input segments \(X\in\mathbb{R}^{N\times d_{model}}\) as follows.
๋ฐ”๋‹๋ผ MHA์˜ ๋‹จ์ผ ํ—ค๋“œ๋Š” ์ž…๋ ฅ ์‹œํ€€์Šค \(X\in\mathbb{R}^{N\times d_{model}}\)๋กœ๋ถ€ํ„ฐ ์ฃผ์˜ ์ปจํ…์ŠคํŠธ \(A_{dot}\in\mathbb{R}^{N\times d_{value}}\)๋ฅผ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๊ณ„์‚ฐํ•ฉ๋‹ˆ๋‹ค.
๋ฐ”๋‹๋ผ MHA์˜ ๋‹จ์ผ ํ—ค๋“œ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ž…๋ ฅ ์„ธ๊ทธ๋จผํŠธ ์‹œํ€€์Šค \(X\in\mathbb{R}^{N\times d_{model}}\)๋กœ๋ถ€ํ„ฐ ์ฃผ์˜ ์ปจํ…์ŠคํŠธ \(A_{dot}\in\mathbb{R}^{N\times d_{value}}\)๋ฅผ ๊ณ„์‚ฐํ•œ๋‹ค.
We also train a multimodal version of MiniCPM, namely MiniCPM-V. The details of MiniCPM-V will be introduced in a separate paper. Here, we only list a snapshot of the performance in Table 11.
์šฐ๋ฆฌ๋Š” ๋˜ํ•œ MiniCPM์˜ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋ฒ„์ „์ธ MiniCPM-V๋ฅผ ํ›ˆ๋ จ์‹œ์ผฐ์Šต๋‹ˆ๋‹ค. MiniCPM-V์˜ ์„ธ๋ถ€ ์‚ฌํ•ญ์€ ๋ณ„๋„์˜ ๋…ผ๋ฌธ์—์„œ ์†Œ๊ฐœ๋  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ๋Š” Table 11์— ์„ฑ๋Šฅ์˜ ์Šค๋ƒ…์ƒท๋งŒ ๋‚˜์—ดํ•ฉ๋‹ˆ๋‹ค.
๋˜ํ•œ ๋‹ค์ค‘ ๋ชจ๋‹ฌ ๋ฒ„์ „์˜ MiniCPM, ์ฆ‰ MiniCPM-V๋ฅผ ํ›ˆ๋ จํ•ฉ๋‹ˆ๋‹ค. MiniCPM-V์— ๋Œ€ํ•œ ์ž์„ธํ•œ ๋‚ด์šฉ์€ ๋ณ„๋„์˜ ๋…ผ๋ฌธ์—์„œ ์†Œ๊ฐœํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ๋Š” ํ‘œ 11์— ์„ฑ๋Šฅ์˜ ์Šค๋ƒ…์ƒท๋งŒ ๋‚˜์—ดํ•ฉ๋‹ˆ๋‹ค.
Preprocess the data: The data needs to be preprocessed to remove noise and irrelevant information. This step is crucial to ensure that the model learns the correct patterns and relationships.
๋ฐ์ดํ„ฐ ์ „์ฒ˜๋ฆฌ: ๋ฐ์ดํ„ฐ์—์„œ ๋…ธ์ด์ฆˆ์™€ ๊ด€๋ จ ์—†๋Š” ์ •๋ณด๋ฅผ ์ œ๊ฑฐํ•˜๊ธฐ ์œ„ํ•ด ์ „์ฒ˜๋ฆฌ๊ฐ€ ํ•„์š”ํ•ฉ๋‹ˆ๋‹ค. ์ด ๋‹จ๊ณ„๋Š” ๋ชจ๋ธ์ด ์˜ฌ๋ฐ”๋ฅธ ํŒจํ„ด๊ณผ ๊ด€๊ณ„๋ฅผ ํ•™์Šตํ•˜๋„๋ก ๋ณด์žฅํ•˜๋Š” ๋ฐ ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค.
๋ฐ์ดํ„ฐ ์ „์ฒ˜๋ฆฌ: ๋ฐ์ดํ„ฐ๋ฅผ ์ „์ฒ˜๋ฆฌํ•˜์—ฌ ๋…ธ์ด์ฆˆ์™€ ๊ด€๋ จ ์—†๋Š” ์ •๋ณด๋ฅผ ์ œ๊ฑฐํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ์ด ๋‹จ๊ณ„๋Š” ๋ชจ๋ธ์ด ์˜ฌ๋ฐ”๋ฅธ ํŒจํ„ด๊ณผ ๊ด€๊ณ„๋ฅผ ํ•™์Šตํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•˜๋Š” ๋ฐ ๋งค์šฐ ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค.
Since SFT and RLHF are two major adaptation tuning methods for LLMs, it is important to understand the connections and difference between them. Next, we make some discussions on this issue30.
SFT์™€ RLHF๋Š” LLMs์˜ ๋‘ ๊ฐ€์ง€ ์ฃผ์š” ์ ์‘ ํŠœ๋‹ ๋ฐฉ๋ฒ•์ด๊ธฐ ๋•Œ๋ฌธ์—, ์ด๋“ค ๊ฐ„์˜ ์—ฐ๊ฒฐ์„ฑ๊ณผ ์ฐจ์ด์ ์„ ์ดํ•ดํ•˜๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค. ๋‹ค์Œ์œผ๋กœ, ์ด ๋ฌธ์ œ์— ๋Œ€ํ•ด ๋ช‡ ๊ฐ€์ง€ ๋…ผ์˜๋ฅผ ์ง„ํ–‰ํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.
SFT์™€ RLHF๋Š” LLM์˜ ์ฃผ์š” ์ ์‘ ํŠœ๋‹ ๋ฐฉ๋ฒ•์ด๋ฏ€๋กœ, ์ด ๋‘˜ ์‚ฌ์ด์˜ ์—ฐ๊ด€์„ฑ๊ณผ ์ฐจ์ด์ ์„ ์ดํ•ดํ•˜๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค. ๋‹ค์Œ์œผ๋กœ, ์ด ๋ฌธ์ œ์— ๋Œ€ํ•ด ๋ช‡ ๊ฐ€์ง€ ๋…ผ์˜๋ฅผ ์ง„ํ–‰ํ•ฉ๋‹ˆ๋‹ค.
Since their salaries increased simultaneously, Lylah's salary also increased by 40% of what she was making four years ago. So, we need to add 40% of $28000 to her salary from four years ago.
๊ทธ๋“ค์˜ ๊ธ‰์—ฌ๊ฐ€ ๋™์‹œ์— ์ธ์ƒ๋˜์—ˆ๊ธฐ ๋•Œ๋ฌธ์—, Lylah์˜ ๊ธ‰์—ฌ๋„ 4๋…„ ์ „ ๊ทธ๋…€๊ฐ€ ๋ฐ›๋˜ ๊ธˆ์•ก์˜ 40%๋งŒํผ ์ฆ๊ฐ€ํ–ˆ์Šต๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ, 4๋…„ ์ „ ๊ทธ๋…€์˜ ๊ธ‰์—ฌ์— $28000์˜ 40%๋ฅผ ๋”ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
๋‘ ์‚ฌ๋žŒ์˜ ๊ธ‰์—ฌ๊ฐ€ ๋™์‹œ์— ์ธ์ƒ๋˜์—ˆ์œผ๋ฏ€๋กœ ๋ฆฌ๋ผ์˜ ๊ธ‰์—ฌ๋„ 4๋…„ ์ „ ๊ธ‰์—ฌ์˜ 40%๊ฐ€ ์ธ์ƒ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ 4๋…„ ์ „ ๊ธ‰์—ฌ์— 28000๋‹ฌ๋Ÿฌ์˜ 40%๋ฅผ ๋”ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
Figure. 2 (Right) shows the results of models using a mixture of GSM8K and MATH in a 1:1 ratio. These models exhibit balanced scaling behaviors in both in-domain and out-of-domain benchmarks.
Figure. 2 (์˜ค๋ฅธ์ชฝ)์€ GSM8K์™€ MATH๋ฅผ 1:1 ๋น„์œจ๋กœ ํ˜ผํ•ฉํ•˜์—ฌ ์‚ฌ์šฉํ•œ ๋ชจ๋ธ์˜ ๊ฒฐ๊ณผ๋ฅผ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค. ์ด ๋ชจ๋ธ๋“ค์€ ๋„๋ฉ”์ธ ๋‚ด ๋ฐ ๋„๋ฉ”์ธ ์™ธ ๋ฒค์น˜๋งˆํฌ์—์„œ ๊ท ํ˜• ์žกํžŒ ํ™•์žฅ ํ–‰๋™์„ ๋‚˜ํƒ€๋ƒ…๋‹ˆ๋‹ค.
๊ทธ๋ฆผ 2 (์˜ค๋ฅธ์ชฝ)๋Š” GSM8K์™€ MATH๋ฅผ 1:1 ๋น„์œจ๋กœ ํ˜ผํ•ฉํ•˜์—ฌ ์‚ฌ์šฉํ•œ ๋ชจ๋ธ์˜ ๊ฒฐ๊ณผ๋ฅผ ๋ณด์—ฌ์ค€๋‹ค. ์ด๋Ÿฌํ•œ ๋ชจ๋ธ์€ ๋„๋ฉ”์ธ ๋‚ด ๋ฐ ๋„๋ฉ”์ธ ์™ธ ๋ฒค์น˜๋งˆํฌ์—์„œ ๊ท ํ˜• ์žกํžŒ ์Šค์ผ€์ผ๋ง ๋™์ž‘์„ ๋ณด์—ฌ์ค€๋‹ค.
**QUESTION:** Jim spends 2 hours watching TV and then decides to go to bed and reads for half as long. He does this 3 times a week. How many hours does he spend on TV and reading in 4 weeks?
**์งˆ๋ฌธ:** Jim์€ TV๋ฅผ 2์‹œ๊ฐ„ ์‹œ์ฒญํ•œ ํ›„ ์ž ์ž๋ฆฌ์— ๋“ค๊ธฐ๋กœ ํ•˜๊ณ  ๊ทธ ์ ˆ๋ฐ˜์˜ ์‹œ๊ฐ„ ๋™์•ˆ ๋…์„œ๋ฅผ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋Š” ์ด๊ฒƒ์„ ์ผ์ฃผ์ผ์— 3๋ฒˆ ํ•ฉ๋‹ˆ๋‹ค. 4์ฃผ ๋™์•ˆ TV ์‹œ์ฒญ๊ณผ ๋…์„œ์— ๋ช‡ ์‹œ๊ฐ„์„ ์†Œ๋น„ํ•ฉ๋‹ˆ๊นŒ?
**์งˆ๋ฌธ:** ์ง์€ TV๋ฅผ ์‹œ์ฒญํ•˜๋Š” ๋ฐ 2์‹œ๊ฐ„์„ ๋ณด๋‚ธ ํ›„ ์ž ์ž๋ฆฌ์— ๋“ค๊ธฐ๋กœ ๊ฒฐ์ •ํ•˜๊ณ  ์ ˆ๋ฐ˜ ์ •๋„๋งŒ ์ฝ์Šต๋‹ˆ๋‹ค. ๊ทธ๋Š” ์ผ์ฃผ์ผ์— ์„ธ ๋ฒˆ ์ด๋ ‡๊ฒŒ ํ•ฉ๋‹ˆ๋‹ค. 4์ฃผ ๋™์•ˆ ์ง์€ TV์™€ ๋…์„œ์— ๋ช‡ ์‹œ๊ฐ„์„ ํ• ์• ํ•˜๋‚˜์š”?
For base models, we consider both common language models, i.e., LLaMA-2 7B/13B/70B/Mistral-7B, and math-specific models, such as Llemma-7B, to assess the generality of the proposed approach.
๊ธฐ๋ณธ ๋ชจ๋ธ๋กœ๋Š” ์ผ๋ฐ˜์ ์ธ ์–ธ์–ด ๋ชจ๋ธ์ธ LLaMA-2 7B/13B/70B/Mistral-7B์™€ ์ˆ˜ํ•™ ์ „์šฉ ๋ชจ๋ธ์ธ Llemma-7B๋ฅผ ๋ชจ๋‘ ๊ณ ๋ คํ•˜์—ฌ ์ œ์•ˆ๋œ ์ ‘๊ทผ ๋ฐฉ์‹์˜ ์ผ๋ฐ˜์„ฑ์„ ํ‰๊ฐ€ํ•ฉ๋‹ˆ๋‹ค.
๊ธฐ๋ณธ ๋ชจ๋ธ์˜ ๊ฒฝ์šฐ, ๊ณตํ†ต ์–ธ์–ด ๋ชจ๋ธ(์˜ˆ: LLaMA-2 7B/13B/70B/Mistral-7B)๊ณผ ์ˆ˜ํ•™ ์ „๋ฌธ ๋ชจ๋ธ(์˜ˆ: Llemma-7B)์„ ๋ชจ๋‘ ๊ณ ๋ คํ•˜์—ฌ ์ œ์•ˆ๋œ ์ ‘๊ทผ ๋ฐฉ์‹์˜ ์ผ๋ฐ˜์„ฑ์„ ํ‰๊ฐ€ํ•œ๋‹ค.
ically, the Yi-6B model's superior performance against models with significantly more parameters challenges the notion that parameter count alone is a sufficient predictor of model efficacy.
ํŠนํžˆ, Yi-6B ๋ชจ๋ธ์˜ ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์€ ๋งค๊ฐœ๋ณ€์ˆ˜ ์ˆ˜๋งŒ์œผ๋กœ ๋ชจ๋ธ์˜ ํšจ์œจ์„ฑ์„ ์ถฉ๋ถ„ํžˆ ์˜ˆ์ธกํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ๊ฐœ๋…์— ๋„์ „ํ•ฉ๋‹ˆ๋‹ค.
๊ธฐ์ˆ ์ ์œผ๋กœ, ๋งค๊ฐœ๋ณ€์ˆ˜๊ฐ€ ํ›จ์”ฌ ๋” ๋งŽ์€ ๋ชจ๋ธ์— ๋Œ€ํ•œ Yi-6B ๋ชจ๋ธ์˜ ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ์€ ๋งค๊ฐœ๋ณ€์ˆ˜ ์ˆ˜๋งŒ์œผ๋กœ ๋ชจ๋ธ ํšจ๋Šฅ์„ ์˜ˆ์ธกํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ๊ฐœ๋…์— ๋„์ „ํ•œ๋‹ค.
In this section, we apply the PLND method to selected languages and models in order to confirm the existence of language-specific neurons and investigate the relationships between languages.
์ด ์„น์…˜์—์„œ๋Š” PLND ๋ฐฉ๋ฒ•์„ ์„ ํƒ๋œ ์–ธ์–ด์™€ ๋ชจ๋ธ์— ์ ์šฉํ•˜์—ฌ ์–ธ์–ด๋ณ„ ๋‰ด๋Ÿฐ์˜ ์กด์žฌ๋ฅผ ํ™•์ธํ•˜๊ณ  ์–ธ์–ด ๊ฐ„์˜ ๊ด€๊ณ„๋ฅผ ์กฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค.
์ด ์„น์…˜์—์„œ๋Š” ์–ธ์–ด๋ณ„ ๋‰ด๋Ÿฐ์˜ ์กด์žฌ๋ฅผ ํ™•์ธํ•˜๊ณ  ์–ธ์–ด ๊ฐ„์˜ ๊ด€๊ณ„๋ฅผ ์กฐ์‚ฌํ•˜๊ธฐ ์œ„ํ•ด PLND ๋ฐฉ๋ฒ•์„ ์„ ํƒ๋œ ์–ธ์–ด์™€ ๋ชจ๋ธ์— ์ ์šฉํ•œ๋‹ค.
- Given his history of strong financial support and political influence, McCarthy may feel compelled to stay and influence the political process, especially with the upcoming 2024 elections.
- ๊ฐ•๋ ฅํ•œ ์žฌ์ • ์ง€์›๊ณผ ์ •์น˜์  ์˜ํ–ฅ๋ ฅ์„ ๊ฐ€์ง„ ๊ทธ์˜ ์ด๋ ฅ์„ ๊ณ ๋ คํ•  ๋•Œ, McCarthy๋Š” ํŠนํžˆ ๋‹ค๊ฐ€์˜ค๋Š” 2024๋…„ ์„ ๊ฑฐ๋ฅผ ์•ž๋‘๊ณ  ์ •์น˜ ๊ณผ์ •์— ๋‚จ์•„ ์˜ํ–ฅ์„ ๋ฏธ์ณ์•ผ ํ•œ๋‹ค๋Š” ์••๋ฐ•์„ ๋Š๋‚„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
- ๊ฐ•๋ ฅํ•œ ์žฌ์ •์  ์ง€์›๊ณผ ์ •์น˜์  ์˜ํ–ฅ๋ ฅ์„ ๋ฐœํœ˜ํ•ด ์˜จ ๊ทธ์˜ ์—ญ์‚ฌ๋ฅผ ๊ณ ๋ คํ•  ๋•Œ, ๋งค์นด์‹œ๋Š” ํŠนํžˆ 2024๋…„ ๋Œ€์„ ์„ ์•ž๋‘๊ณ  ์ •์น˜ ๊ณผ์ •์— ๊ณ„์† ์ฐธ์—ฌํ•˜๊ณ  ์˜ํ–ฅ๋ ฅ์„ ํ–‰์‚ฌํ•ด์•ผ ํ•œ๋‹ค๊ณ  ๋Š๋‚„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
Xiao Wang,, Shiao Wang, Yuhe Ding, Yuehang Li, Wentao Wu, Yao Rong, Weizhe Kong, Ju Huang, Shihao Li, Haoxiang Yang, Ziwen Wang, Bo Jiang, Chenglong Li, Yaowei Wang, Yonghong Tian,, Jin Tang
Xiao Wang, Shiao Wang, Yuhe Ding, Yuehang Li, Wentao Wu, Yao Rong, Weizhe Kong, Ju Huang, Shihao Li, Haoxiang Yang, Ziwen Wang, Bo Jiang, Chenglong Li, Yaowei Wang, Yonghong Tian, Jin Tang
์ƒค์˜ค ์™•, ์‹œ์•„์˜ค ์™•, ์œ ํ—ˆ ๋”ฉ, ์œ„ํ•œ ๋ฆฌ, ์›ํƒ€์˜ค ์šฐ, ์•ผ์˜ค ๋ฃฝ, ์œ„์ œ ์ฝฉ, ์ฃผ ํ™ฉ, ์‹œํ•˜์˜ค ๋ฆฌ, ํ•˜์˜ค์ƒน ์–‘, ์ง€์›ฌ ์™•, ๋ณด ์žฅ, ์ฒธ๋ฃฝ ๋ฆฌ, ์•ผ์˜ค์›จ์ด ์™•, ์šฉํ™ ํ‹ฐ์•ˆ, ์ง„ํƒ•
This section will introduce some of the data sources commonly used in RALM and categorise them into structured and unstructured data. Figure 5 illustrates the categorization of data sources.
์ด ์„น์…˜์—์„œ๋Š” RALM์—์„œ ์ผ๋ฐ˜์ ์œผ๋กœ ์‚ฌ์šฉ๋˜๋Š” ๋ฐ์ดํ„ฐ ์†Œ์Šค ๋ช‡ ๊ฐ€์ง€๋ฅผ ์†Œ๊ฐœํ•˜๊ณ  ์ด๋ฅผ ๊ตฌ์กฐํ™”๋œ ๋ฐ์ดํ„ฐ์™€ ๋น„๊ตฌ์กฐํ™”๋œ ๋ฐ์ดํ„ฐ๋กœ ๋ถ„๋ฅ˜ํ•ฉ๋‹ˆ๋‹ค. Figure 5๋Š” ๋ฐ์ดํ„ฐ ์†Œ์Šค์˜ ๋ถ„๋ฅ˜๋ฅผ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
์ด ์„น์…˜์—์„œ๋Š” RALM์—์„œ ์ผ๋ฐ˜์ ์œผ๋กœ ์‚ฌ์šฉ๋˜๋Š” ๋ช‡ ๊ฐ€์ง€ ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ฅผ ์†Œ๊ฐœํ•˜๊ณ  ์ด๋ฅผ ์ •ํ˜• ๋ฐ์ดํ„ฐ์™€ ๋น„์ •ํ˜• ๋ฐ์ดํ„ฐ๋กœ ๋ถ„๋ฅ˜ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ฆผ 5๋Š” ๋ฐ์ดํ„ฐ ์†Œ์Šค์˜ ๋ถ„๋ฅ˜๋ฅผ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
We use "Answer the following question with a short span, or a full and complete answer." for SQA and TopiOCQA, since these datasets have both short and long answers based on the questions.
SQA์™€ TopiOCQA์— ๋Œ€ํ•ด์„œ๋Š” "๋‹ค์Œ ์งˆ๋ฌธ์— ์งง์€ ๋‹ต๋ณ€ ๋˜๋Š” ์™„์ „ํ•˜๊ณ  ์™„๋ฒฝํ•œ ๋‹ต๋ณ€์œผ๋กœ ๋‹ตํ•˜์‹œ์˜ค."๋ฅผ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค. ์ด ๋ฐ์ดํ„ฐ์…‹๋“ค์€ ์งˆ๋ฌธ์— ๋”ฐ๋ผ ์งง์€ ๋‹ต๋ณ€๊ณผ ๊ธด ๋‹ต๋ณ€์„ ๋ชจ๋‘ ํฌํ•จํ•˜๊ณ  ์žˆ๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค.
์ด๋Ÿฌํ•œ ๋ฐ์ดํ„ฐ ์„ธํŠธ์—๋Š” ์งˆ๋ฌธ์— ๋”ฐ๋ผ ์งง์€ ๋‹ต๋ณ€๊ณผ ๊ธด ๋‹ต๋ณ€์ด ๋ชจ๋‘ ์žˆ๊ธฐ ๋•Œ๋ฌธ์— SQA์™€ TopiOCQA์—๋Š” "์งง์€ ๋ฒ”์œ„ ๋˜๋Š” ์™„์ „ํ•˜๊ณ  ์™„์ „ํ•œ ๋‹ต๋ณ€์œผ๋กœ ๋‹ค์Œ ์งˆ๋ฌธ์— ๋‹ตํ•˜์„ธ์š”."๋ฅผ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.
Substituting \(S_{min}\) with \(S\)Finally, after establishing the analytical relationship between \(S_{min}\) and \(S\), we can substitute \(S_{min}\) with \(S\) to link \(S\) with \(L\).
๋งˆ์ง€๋ง‰์œผ๋กœ, \(S_{min}\)๊ณผ \(S\) ์‚ฌ์ด์˜ ๋ถ„์„์  ๊ด€๊ณ„๋ฅผ ํ™•๋ฆฝํ•œ ํ›„, \(S_{min}\)์„ \(S\)๋กœ ๋Œ€์ฒดํ•˜์—ฌ \(S\)์™€ \(L\)์„ ์—ฐ๊ฒฐํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
S_{min}\)์„ \(S\)๋กœ ๋Œ€์ž…๋งˆ์ง€๋ง‰์œผ๋กœ, \(S_{min}\)๊ณผ \(S\) ์‚ฌ์ด์˜ ๋ถ„์„์  ๊ด€๊ณ„๋ฅผ ์„ค์ •ํ•œ ํ›„, \(S_{min}\)์„ \(S\)๋กœ ๋Œ€์ฒดํ•˜์—ฌ \(S\)์™€ \(L\)์„ ์—ฐ๊ฒฐํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
In our study, since the preferred data is selected by reference-free models and the same models are used for evaluation, we investigate the potential for 'cheating' in the scoring process.
์šฐ๋ฆฌ์˜ ์—ฐ๊ตฌ์—์„œ, ์„ ํ˜ธ๋˜๋Š” ๋ฐ์ดํ„ฐ๋Š” ์ฐธ์กฐ ์—†๋Š” ๋ชจ๋ธ๋“ค์— ์˜ํ•ด ์„ ํƒ๋˜๊ณ  ๋™์ผํ•œ ๋ชจ๋ธ๋“ค์ด ํ‰๊ฐ€์— ์‚ฌ์šฉ๋˜๊ธฐ ๋•Œ๋ฌธ์—, ์šฐ๋ฆฌ๋Š” ์ ์ˆ˜ ๋งค๊ธฐ๊ธฐ ๊ณผ์ •์—์„œ '๋ถ€์ •ํ–‰์œ„'์˜ ๊ฐ€๋Šฅ์„ฑ์„ ์กฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค.
๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ์ฐธ์กฐ ์—†๋Š” ๋ชจ๋ธ์— ์˜ํ•ด ์„ ํ˜ธ ๋ฐ์ดํ„ฐ๊ฐ€ ์„ ํƒ๋˜๊ณ  ๋™์ผํ•œ ๋ชจ๋ธ์ด ํ‰๊ฐ€์— ์‚ฌ์šฉ๋˜๊ธฐ ๋•Œ๋ฌธ์— ์ฑ„์  ๊ณผ์ •์—์„œ '๋ถ€์ •ํ–‰์œ„'์˜ ๊ฐ€๋Šฅ์„ฑ์„ ์กฐ์‚ฌํ•œ๋‹ค.
These enhancement techniques lead to the improved GPT-3 models with stronger capacities, which are called GPT-3.5 models by OpenAI (see the discussion about the OpenAI API in Section 3.1).
์ด๋Ÿฌํ•œ ํ–ฅ์ƒ ๊ธฐ์ˆ ๋“ค์€ ๋” ๊ฐ•๋ ฅํ•œ ๋Šฅ๋ ฅ์„ ๊ฐ€์ง„ ๊ฐœ์„ ๋œ GPT-3 ๋ชจ๋ธ๋กœ ์ด์–ด์ง€๋ฉฐ, ์ด๋Š” OpenAI์— ์˜ํ•ด GPT-3.5 ๋ชจ๋ธ๋กœ ๋ถˆ๋ฆฝ๋‹ˆ๋‹ค (OpenAI API์— ๋Œ€ํ•œ ๋…ผ์˜๋Š” 3.1์ ˆ์„ ์ฐธ์กฐํ•˜์‹ญ์‹œ์˜ค).
์ด๋Ÿฌํ•œ ํ–ฅ์ƒ ๊ธฐ๋ฒ•์€ ๋” ๊ฐ•๋ ฅํ•œ ์šฉ๋Ÿ‰์„ ๊ฐ–์ถ˜ ๊ฐœ์„ ๋œ GPT-3 ๋ชจ๋ธ๋กœ ์ด์–ด์ง€๋ฉฐ, ์ด๋ฅผ OpenAI์—์„œ๋Š” GPT-3.5 ๋ชจ๋ธ์ด๋ผ๊ณ  ๋ถ€๋ฆ…๋‹ˆ๋‹ค(์„น์…˜ 3.1์˜ OpenAI API์— ๋Œ€ํ•œ ๋…ผ์˜ ์ฐธ์กฐ).
2. Recovery: HIIT involves short recovery periods, which can help to increase muscle strength and endurance. Other forms of exercise, such as jogging, may not provide enough recovery time.
2. ํšŒ๋ณต: HIIT๋Š” ์งง์€ ํšŒ๋ณต ๊ธฐ๊ฐ„์„ ํฌํ•จํ•˜๋ฉฐ, ์ด๋Š” ๊ทผ๋ ฅ๊ณผ ์ง€๊ตฌ๋ ฅ์„ ์ฆ๊ฐ€์‹œํ‚ค๋Š” ๋ฐ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์กฐ๊น…๊ณผ ๊ฐ™์€ ๋‹ค๋ฅธ ํ˜•ํƒœ์˜ ์šด๋™์€ ์ถฉ๋ถ„ํ•œ ํšŒ๋ณต ์‹œ๊ฐ„์„ ์ œ๊ณตํ•˜์ง€ ์•Š์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
2. ํšŒ๋ณต: HIIT๋Š” ์งง์€ ํšŒ๋ณต ๊ธฐ๊ฐ„์„ ํฌํ•จํ•˜๋ฏ€๋กœ ๊ทผ๋ ฅ๊ณผ ์ง€๊ตฌ๋ ฅ์„ ํ–ฅ์ƒ์‹œํ‚ค๋Š” ๋ฐ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์กฐ๊น…๊ณผ ๊ฐ™์€ ๋‹ค๋ฅธ ํ˜•ํƒœ์˜ ์šด๋™์€ ์ถฉ๋ถ„ํ•œ ํšŒ๋ณต ์‹œ๊ฐ„์„ ์ œ๊ณตํ•˜์ง€ ๋ชปํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
Figure 4: _Naive_ and _overall_ efficiency gain of data selection via D4 relative to the total cost of training as a function of model size on Instruct + Answers perplexity at \(R=0.25\).
Figure 4: Instruct + Answers ๋‹นํ˜น๋„์—์„œ \(R=0.25\)์ผ ๋•Œ ๋ชจ๋ธ ํฌ๊ธฐ์˜ ํ•จ์ˆ˜๋กœ์„œ D4๋ฅผ ํ†ตํ•œ ๋ฐ์ดํ„ฐ ์„ ํƒ์˜ _Naive_ ๋ฐ _overall_ ํšจ์œจ์„ฑ ์ฆ๊ฐ€๋ฅผ ์ „์ฒด ํ›ˆ๋ จ ๋น„์šฉ๊ณผ ๋น„๊ตํ•œ ๊ฒฐ๊ณผ.
๊ทธ๋ฆผ 4: Instruct + Answers ๋‚œํ•ด๋„์—์„œ ๋ชจ๋ธ ํฌ๊ธฐ์— ๋Œ€ํ•œ ํ•จ์ˆ˜๋กœ์„œ D4๋ฅผ ํ†ตํ•œ ๋ฐ์ดํ„ฐ ์„ ํƒ์˜ _์ˆœ์ง„ํ•œ_ ํšจ์œจ์„ฑ๊ณผ ์ด ํ›ˆ๋ จ ๋น„์šฉ์— ๋Œ€ํ•œ _์ „์ฒด์ ์ธ_ ํšจ์œจ์„ฑ ์ด๋“(\(R=0.25\)).
Figure 12: Same as Figure 2, but including models with accuracies below 50% (which may overlap with higher-accuracy models). The _peak_ capacity ratios consistently exceed \(R(F)\geq 2\).
Figure 12: Figure 2์™€ ๋™์ผํ•˜์ง€๋งŒ, ์ •ํ™•๋„๊ฐ€ 50% ์ดํ•˜์ธ ๋ชจ๋ธ๋„ ํฌํ•จ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค (์ด๋Š” ๋” ๋†’์€ ์ •ํ™•๋„์˜ ๋ชจ๋ธ๊ณผ ๊ฒน์น  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค). _์ตœ๋Œ€_ ์šฉ๋Ÿ‰ ๋น„์œจ์€ ์ผ๊ด€๋˜๊ฒŒ \(R(F)\geq 2\)๋ฅผ ์ดˆ๊ณผํ•ฉ๋‹ˆ๋‹ค.
๊ทธ๋ฆผ 12: ๊ทธ๋ฆผ 2์™€ ๋™์ผํ•˜์ง€๋งŒ ์ •ํ™•๋„๊ฐ€ 50% ๋ฏธ๋งŒ์ธ ๋ชจ๋ธ(์ •ํ™•๋„๊ฐ€ ๋†’์€ ๋ชจ๋ธ๊ณผ ๊ฒน์น  ์ˆ˜ ์žˆ์Œ)์„ ํฌํ•จํ•ฉ๋‹ˆ๋‹ค. ํ”ผํฌ_ ์šฉ๋Ÿ‰ ๋น„์œจ์€ ์ผ๊ด€๋˜๊ฒŒ \(R(F)\geq 2\)๋ฅผ ์ดˆ๊ณผํ•ฉ๋‹ˆ๋‹ค.
Footnote 2: Most duplicates we saw were automatically generated pages, such as the outcomes of sports games. This shows the strength of manual curation for creating high-quality datasets.
Footnote 2: ์šฐ๋ฆฌ๊ฐ€ ๋ณธ ๋Œ€๋ถ€๋ถ„์˜ ์ค‘๋ณต ํŽ˜์ด์ง€๋Š” ์Šคํฌ์ธ  ๊ฒฝ๊ธฐ ๊ฒฐ๊ณผ์™€ ๊ฐ™์€ ์ž๋™ ์ƒ์„ฑ๋œ ํŽ˜์ด์ง€์˜€์Šต๋‹ˆ๋‹ค. ์ด๋Š” ๊ณ ํ’ˆ์งˆ ๋ฐ์ดํ„ฐ์…‹์„ ๋งŒ๋“ค๊ธฐ ์œ„ํ•œ ์ˆ˜์ž‘์—… ํ๋ ˆ์ด์…˜์˜ ๊ฐ•์ ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
๊ฐ์ฃผ 2: ์šฐ๋ฆฌ๊ฐ€ ๋ณธ ๋Œ€๋ถ€๋ถ„์˜ ์ค‘๋ณต์€ ์Šคํฌ์ธ  ๊ฒฝ๊ธฐ ๊ฒฐ๊ณผ์™€ ๊ฐ™์ด ์ž๋™์œผ๋กœ ์ƒ์„ฑ๋œ ํŽ˜์ด์ง€์˜€์Šต๋‹ˆ๋‹ค. ์ด๋Š” ๊ณ ํ’ˆ์งˆ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ๋งŒ๋“œ๋Š” ๋ฐ ์žˆ์–ด ์ˆ˜๋™ ํ๋ ˆ์ด์…˜์˜ ๊ฐ•์ ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
Some of the real-world datasets used in the development of AMIE are open-source (MedQA). The scenario packs from UK used in the OSCE study are also available for download on the internet.
AMIE ๊ฐœ๋ฐœ์— ์‚ฌ์šฉ๋œ ์ผ๋ถ€ ์‹ค์ œ ๋ฐ์ดํ„ฐ์…‹์€ ์˜คํ”ˆ ์†Œ์Šค์ž…๋‹ˆ๋‹ค (MedQA). OSCE ์—ฐ๊ตฌ์— ์‚ฌ์šฉ๋œ ์˜๊ตญ์˜ ์‹œ๋‚˜๋ฆฌ์˜ค ํŒฉ๋„ ์ธํ„ฐ๋„ท์—์„œ ๋‹ค์šด๋กœ๋“œํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
AMIE ๊ฐœ๋ฐœ์— ์‚ฌ์šฉ๋œ ์‹ค์ œ ๋ฐ์ดํ„ฐ ์„ธํŠธ ์ค‘ ์ผ๋ถ€๋Š” ์˜คํ”ˆ ์†Œ์Šค(MedQA)์ž…๋‹ˆ๋‹ค. OSCE ์—ฐ๊ตฌ์— ์‚ฌ์šฉ๋œ ์˜๊ตญ์˜ ์‹œ๋‚˜๋ฆฌ์˜ค ํŒฉ๋„ ์ธํ„ฐ๋„ท์—์„œ ๋‹ค์šด๋กœ๋“œํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
For the stage-1 SFT data used in stage-2 instruction tuning, we keep the same format template as in stage-1. In other words, the {Context for Latest Question} and {Instruction} are empty.
2๋‹จ๊ณ„ instruction tuning์— ์‚ฌ์šฉ๋˜๋Š” 1๋‹จ๊ณ„ SFT ๋ฐ์ดํ„ฐ์˜ ๊ฒฝ์šฐ, 1๋‹จ๊ณ„์™€ ๋™์ผํ•œ ํ˜•์‹ ํ…œํ”Œ๋ฆฟ์„ ์œ ์ง€ํ•ฉ๋‹ˆ๋‹ค. ์ฆ‰, {Context for Latest Question}์™€ {Instruction}์€ ๋น„์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
2๋‹จ๊ณ„ ์ง€์นจ ํŠœ๋‹์— ์‚ฌ์šฉ๋˜๋Š” 1๋‹จ๊ณ„ SFT ๋ฐ์ดํ„ฐ์˜ ๊ฒฝ์šฐ, 1๋‹จ๊ณ„์™€ ๋™์ผํ•œ ํ˜•์‹ ํ…œํ”Œ๋ฆฟ์„ ์œ ์ง€ํ•ฉ๋‹ˆ๋‹ค. ์ฆ‰, '{์ตœ์‹  ์งˆ๋ฌธ์— ๋Œ€ํ•œ ์ปจํ…์ŠคํŠธ}'์™€ '{์ง€์นจ}'์€ ๋น„์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
Finally, to demonstrate why it is reasonable to assume that models cannot generally answer questions about new information, with better than random success, look at the following example:
๋งˆ์ง€๋ง‰์œผ๋กœ, ๋ชจ๋ธ์ด ์ผ๋ฐ˜์ ์œผ๋กœ ์ƒˆ๋กœ์šด ์ •๋ณด์— ๋Œ€ํ•œ ์งˆ๋ฌธ์— ๋ฌด์ž‘์œ„ ์„ฑ๊ณต๋ฅ ๋ณด๋‹ค ๋” ๋‚˜์€ ์„ฑ๊ณผ๋กœ ๋‹ตํ•  ์ˆ˜ ์—†๋‹ค๊ณ  ๊ฐ€์ •ํ•˜๋Š” ๊ฒƒ์ด ์™œ ํ•ฉ๋ฆฌ์ ์ธ์ง€ ์„ค๋ช…ํ•˜๊ธฐ ์œ„ํ•ด, ๋‹ค์Œ ์˜ˆ๋ฅผ ์‚ดํŽด๋ณด์‹ญ์‹œ์˜ค:
๋งˆ์ง€๋ง‰์œผ๋กœ, ๋ชจ๋ธ์ด ์ผ๋ฐ˜์ ์œผ๋กœ ์ƒˆ๋กœ์šด ์ •๋ณด์— ๋Œ€ํ•œ ์งˆ๋ฌธ์— ๋ฌด์ž‘์œ„๋ณด๋‹ค ๋” ๋‚˜์€ ์„ฑ๊ณต๋ฅ ๋กœ ๋Œ€๋‹ตํ•  ์ˆ˜ ์—†๋‹ค๊ณ  ๊ฐ€์ •ํ•˜๋Š” ๊ฒƒ์ด ํ•ฉ๋ฆฌ์ ์ธ ์ด์œ ๋ฅผ ๋ณด์—ฌ์ฃผ๊ธฐ ์œ„ํ•ด ๋‹ค์Œ ์˜ˆ๋ฅผ ์‚ดํŽด๋ณด์„ธ์š”:
detailed information about access requests, agent activities, and any modifications to the access control parameters, which help to safeguard against potential privilege attacks [64; 65].
์ ‘๊ทผ ์š”์ฒญ, ์—์ด์ „ํŠธ ํ™œ๋™ ๋ฐ ์ ‘๊ทผ ์ œ์–ด ๋งค๊ฐœ๋ณ€์ˆ˜์˜ ๋ชจ๋“  ์ˆ˜์ • ์‚ฌํ•ญ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ์ •๋ณด๋Š” ์ž ์žฌ์ ์ธ ๊ถŒํ•œ ๊ณต๊ฒฉ์„ ๋ฐฉ์ง€ํ•˜๋Š” ๋ฐ ๋„์›€์ด ๋ฉ๋‹ˆ๋‹ค [64; 65].
์•ก์„ธ์Šค ์š”์ฒญ, ์—์ด์ „ํŠธ ํ™œ๋™ ๋ฐ ์•ก์„ธ์Šค ์ œ์–ด ๋งค๊ฐœ๋ณ€์ˆ˜์˜ ๋ชจ๋“  ์ˆ˜์ • ์‚ฌํ•ญ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ์ •๋ณด๋Š” ์ž ์žฌ์ ์ธ ๊ถŒํ•œ ๊ณต๊ฒฉ์œผ๋กœ๋ถ€ํ„ฐ ๋ณดํ˜ธํ•˜๋Š” ๋ฐ ๋„์›€์ด ๋ฉ๋‹ˆ๋‹ค[64; 65].
**Evaluation of Fine-tuned LLMs.** Fine-tuned LLMs in this part refer to the model checkpoints obtained after instruction tuning or alignment tuning based on pre-trained model weights45.
**๋ฏธ์„ธ ์กฐ์ •๋œ LLM์˜ ํ‰๊ฐ€.** ์ด ๋ถ€๋ถ„์—์„œ ๋ฏธ์„ธ ์กฐ์ •๋œ LLM์€ ์‚ฌ์ „ ํ›ˆ๋ จ๋œ ๋ชจ๋ธ ๊ฐ€์ค‘์น˜๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ instruction tuning ๋˜๋Š” alignment tuning ํ›„์— ์–ป์€ ๋ชจ๋ธ ์ฒดํฌํฌ์ธํŠธ๋ฅผ ์˜๋ฏธํ•ฉ๋‹ˆ๋‹ค.
**๋ฏธ์„ธ ์กฐ์ •๋œ LLM์˜ ํ‰๊ฐ€**. ์ด ๋ถ€๋ถ„์—์„œ ๋ฏธ์„ธ ์กฐ์ •๋œ LLM์€ ์‚ฌ์ „ ํ•™์Šต๋œ ๋ชจ๋ธ ๊ฐ€์ค‘์น˜45๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๋ช…๋ น์–ด ํŠœ๋‹ ๋˜๋Š” ์ •๋ ฌ ํŠœ๋‹ ํ›„ ์–ป์€ ๋ชจ๋ธ ์ฒดํฌํฌ์ธํŠธ๋ฅผ ์˜๋ฏธํ•ฉ๋‹ˆ๋‹ค.
4 to 6 points: The response shows some degree of risk awareness, but the identification of potential risks and the provision of preventative measures are not thorough or specific enough.
4์ ์—์„œ 6์ : ์‘๋‹ต์€ ์–ด๋Š ์ •๋„์˜ ์œ„ํ—˜ ์ธ์‹์„ ๋ณด์—ฌ์ฃผ์ง€๋งŒ, ์ž ์žฌ์  ์œ„ํ—˜์˜ ์‹๋ณ„๊ณผ ์˜ˆ๋ฐฉ ์กฐ์น˜์˜ ์ œ๊ณต์ด ์ถฉ๋ถ„ํžˆ ์ฒ ์ €ํ•˜๊ฑฐ๋‚˜ ๊ตฌ์ฒด์ ์ด์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
4~6์ : ์‘๋‹ต์€ ์–ด๋Š ์ •๋„์˜ ์œ„ํ—˜ ์ธ์‹์„ ๋ณด์—ฌ ์ฃผ์ง€๋งŒ ์ž ์žฌ์  ์œ„ํ—˜์„ ์‹๋ณ„ํ•˜๊ณ  ์˜ˆ๋ฐฉ ์กฐ์น˜๋ฅผ ์ œ๊ณตํ•˜๋Š” ๊ฒƒ์ด ์ถฉ๋ถ„ํžˆ ์ฒ ์ €ํ•˜๊ฑฐ๋‚˜ ๊ตฌ์ฒด์ ์ด์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
This adds only a single scalar value as training parameter per head while allowing a learnable trade-off between the long-term and local information flows in the model (Wu et al., 2022).
์ด๋Š” ๊ฐ ํ—ค๋“œ๋‹น ํ•˜๋‚˜์˜ ์Šค์นผ๋ผ ๊ฐ’์„ ํ›ˆ๋ จ ๋งค๊ฐœ๋ณ€์ˆ˜๋กœ ์ถ”๊ฐ€ํ•˜๋ฉด์„œ ๋ชจ๋ธ ๋‚ด์—์„œ ์žฅ๊ธฐ ๋ฐ ๋กœ์ปฌ ์ •๋ณด ํ๋ฆ„ ๊ฐ„์˜ ํ•™์Šต ๊ฐ€๋Šฅํ•œ ๊ท ํ˜•์„ ํ—ˆ์šฉํ•ฉ๋‹ˆ๋‹ค (Wu et al., 2022).
์ด๋Š” ๋ชจ๋ธ์˜ ์žฅ๊ธฐ ๋ฐ ๋กœ์ปฌ ์ •๋ณด ํ๋ฆ„ ๊ฐ„์— ํ•™์Šต ๊ฐ€๋Šฅํ•œ ํŠธ๋ ˆ์ด๋“œ์˜คํ”„๋ฅผ ํ—ˆ์šฉํ•˜๋ฉด์„œ ํ—ค๋“œ๋‹น ๋‹จ์ผ ์Šค์นผ๋ผ ๊ฐ’๋งŒ์„ ํ›ˆ๋ จ ํŒŒ๋ผ๋ฏธํ„ฐ๋กœ ์ถ”๊ฐ€ํ•œ๋‹ค(Wu et al., 2022).
Footnote 2: We omit the bias term as it often models the frequency of the token (Kobayashi et al., 2023), which we do not want to consider for measuring the internal recall of an entity.
Footnote 2: ์šฐ๋ฆฌ๋Š” ํŽธํ–ฅ ํ•ญ์„ ์ƒ๋žตํ•ฉ๋‹ˆ๋‹ค. ์ด๋Š” ์ข…์ข… ํ† ํฐ์˜ ๋นˆ๋„๋ฅผ ๋ชจ๋ธ๋งํ•˜๋Š”๋ฐ ์‚ฌ์šฉ๋˜๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค(Kobayashi et al., 2023). ์ด๋Š” ์—”ํ‹ฐํ‹ฐ์˜ ๋‚ด๋ถ€ ํšŒ์ƒ์„ ์ธก์ •ํ•˜๋Š” ๋ฐ ๊ณ ๋ คํ•˜๊ณ  ์‹ถ์ง€ ์•Š์€ ์š”์†Œ์ž…๋‹ˆ๋‹ค.
๊ฐ์ฃผ 2: ๋ฐ”์ด์–ด์Šค ํ•ญ์€ ์ข…์ข… ํ† ํฐ์˜ ๋นˆ๋„๋ฅผ ๋ชจ๋ธ๋งํ•˜๊ธฐ ๋•Œ๋ฌธ์— ์ƒ๋žตํ•ฉ๋‹ˆ๋‹ค(Kobayashi et al., 2023), ์ด๋Š” ์—”ํ‹ฐํ‹ฐ์˜ ๋‚ด๋ถ€ ๋ฆฌ์ฝœ์„ ์ธก์ •ํ•  ๋•Œ ๊ณ ๋ คํ•˜๊ณ  ์‹ถ์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
clear pitfalls -- for example, if we took thousands of wikipedia articles about unrelated topics and appended the same email at the end of each article, they might be clustered together.
๋ช…ํ™•ํ•œ ํ•จ์ • -- ์˜ˆ๋ฅผ ๋“ค์–ด, ๊ด€๋ จ ์—†๋Š” ์ฃผ์ œ์— ๋Œ€ํ•œ ์ˆ˜์ฒœ ๊ฐœ์˜ ์œ„ํ‚คํ”ผ๋””์•„ ๊ธฐ์‚ฌ๋ฅผ ๊ฐ€์ ธ์™€ ๊ฐ ๊ธฐ์‚ฌ ๋์— ๋™์ผํ•œ ์ด๋ฉ”์ผ์„ ์ฒจ๋ถ€ํ•˜๋ฉด, ์ด๋“ค์€ ํ•จ๊ป˜ ํด๋Ÿฌ์Šคํ„ฐ๋ง๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
์˜ˆ๋ฅผ ๋“ค์–ด, ๊ด€๋ จ ์—†๋Š” ์ฃผ์ œ์— ๋Œ€ํ•œ ์ˆ˜์ฒœ ๊ฐœ์˜ ์œ„ํ‚ค๋ฐฑ๊ณผ ๊ธฐ์‚ฌ๋ฅผ ๊ฐ€์ ธ์™€ ๊ฐ ๊ธฐ์‚ฌ ๋์— ๋™์ผํ•œ ์ด๋ฉ”์ผ์„ ์ถ”๊ฐ€ํ•˜๋ฉด ํด๋Ÿฌ์Šคํ„ฐ๋ง์ด ํ•จ๊ป˜ ์ด๋ฃจ์–ด์งˆ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
This publication is part of the project LESSEN with project number NWA.1389.20.183 of the research program NWA ORC 2020/21 which is (partly) financed by the Dutch Research Council (NWO).
์ด ์ถœํŒ๋ฌผ์€ ๋„ค๋œ๋ž€๋“œ ์—ฐ๊ตฌ ์œ„์›ํšŒ (NWO)๊ฐ€ (๋ถ€๋ถ„์ ์œผ๋กœ) ์ž๊ธˆ์„ ์ง€์›ํ•˜๋Š” ์—ฐ๊ตฌ ํ”„๋กœ๊ทธ๋žจ NWA ORC 2020/21์˜ ํ”„๋กœ์ ํŠธ ๋ฒˆํ˜ธ NWA.1389.20.183์ธ LESSEN ํ”„๋กœ์ ํŠธ์˜ ์ผ๋ถ€์ž…๋‹ˆ๋‹ค.
์ด ์ถœํŒ๋ฌผ์€ ๋„ค๋œ๋ž€๋“œ ์—ฐ๊ตฌ์œ„์›ํšŒ(NWO)์˜ (๋ถ€๋ถ„) ์žฌ์ • ์ง€์›์„ ๋ฐ›๋Š” ์—ฐ๊ตฌ ํ”„๋กœ๊ทธ๋žจ NWA ORC 2020/21์˜ ํ”„๋กœ์ ํŠธ ๋ฒˆํ˜ธ NWA.1389.20.183์„ ๊ฐ€์ง„ LESSEN ํ”„๋กœ์ ํŠธ์˜ ์ผ๋ถ€์ž…๋‹ˆ๋‹ค.
**Knowledge Pieces.** Defining "one piece of human knowledge" precisely is challenging. This paper aims to make progress by focusing on a restricted, yet sufficiently interesting domain.
**์ง€์‹ ์กฐ๊ฐ.** "ํ•˜๋‚˜์˜ ์ธ๊ฐ„ ์ง€์‹ ์กฐ๊ฐ"์„ ์ •ํ™•ํ•˜๊ฒŒ ์ •์˜ํ•˜๋Š” ๊ฒƒ์€ ์–ด๋ ต์Šต๋‹ˆ๋‹ค. ์ด ๋…ผ๋ฌธ์€ ์ œํ•œ์ ์ด์ง€๋งŒ ์ถฉ๋ถ„ํžˆ ํฅ๋ฏธ๋กœ์šด ๋„๋ฉ”์ธ์— ์ดˆ์ ์„ ๋งž์ถ”์–ด ์ง„์ „์„ ์ด๋ฃจ๊ณ ์ž ํ•ฉ๋‹ˆ๋‹ค.
**์ง€์‹ ์กฐ๊ฐ**. "์ธ๊ฐ„ ์ง€์‹์˜ ํ•œ ์กฐ๊ฐ"์„ ์ •ํ™•ํ•˜๊ฒŒ ์ •์˜ํ•˜๋Š” ๊ฒƒ์€ ์–ด๋ ค์šด ์ผ์ž…๋‹ˆ๋‹ค. ์ด ๋ฐฑ์„œ์—์„œ๋Š” ์ œํ•œ์ ์ด์ง€๋งŒ ์ถฉ๋ถ„ํžˆ ํฅ๋ฏธ๋กœ์šด ์˜์—ญ์— ์ดˆ์ ์„ ๋งž์ถ”์–ด ์ง„์ „์„ ์ด๋ฃจ๊ณ ์ž ํ•ฉ๋‹ˆ๋‹ค.
Appendix A shows some examples generated by our _WizardMath_. The examples demonstrate that our model consistently generates accurate response answers accompanied by clear explanations.
๋ถ€๋ก A์—๋Š” _WizardMath_์— ์˜ํ•ด ์ƒ์„ฑ๋œ ๋ช‡ ๊ฐ€์ง€ ์˜ˆ์‹œ๊ฐ€ ๋‚˜์™€ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด ์˜ˆ์‹œ๋“ค์€ ์šฐ๋ฆฌ ๋ชจ๋ธ์ด ๋ช…ํ™•ํ•œ ์„ค๋ช…๊ณผ ํ•จ๊ป˜ ์ผ๊ด€๋˜๊ฒŒ ์ •ํ™•ํ•œ ์‘๋‹ต์„ ์ƒ์„ฑํ•จ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
๋ถ€๋ก A์—๋Š” _WizardMath_์—์„œ ์ƒ์„ฑํ•œ ๋ช‡ ๊ฐ€์ง€ ์˜ˆ๊ฐ€ ๋‚˜์™€ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด ์˜ˆ์ œ๋Š” ์šฐ๋ฆฌ ๋ชจ๋ธ์ด ๋ช…ํ™•ํ•œ ์„ค๋ช…๊ณผ ํ•จ๊ป˜ ์ผ๊ด€๋˜๊ฒŒ ์ •ํ™•ํ•œ ์‘๋‹ต ๋‹ต์„ ์ƒ์„ฑํ•œ๋‹ค๋Š” ๊ฒƒ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
Informativeness: Judges if the answer provides sufficient relevant information of practical value. A high-quality answer should offer useful and pertinent information to the questioner.
์ •๋ณด์„ฑ: ๋‹ต๋ณ€์ด ์‹ค์งˆ์ ์ธ ๊ฐ€์น˜๊ฐ€ ์žˆ๋Š” ์ถฉ๋ถ„ํ•œ ๊ด€๋ จ ์ •๋ณด๋ฅผ ์ œ๊ณตํ•˜๋Š”์ง€ ํŒ๋‹จํ•ฉ๋‹ˆ๋‹ค. ๊ณ ํ’ˆ์งˆ์˜ ๋‹ต๋ณ€์€ ์งˆ๋ฌธ์ž์—๊ฒŒ ์œ ์šฉํ•˜๊ณ  ์ ์ ˆํ•œ ์ •๋ณด๋ฅผ ์ œ๊ณตํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
์œ ์ต์„ฑ: ๋‹ต๋ณ€์ด ์‹ค์งˆ์ ์ธ ๊ฐ€์น˜๊ฐ€ ์žˆ๋Š” ์ถฉ๋ถ„ํ•œ ๊ด€๋ จ ์ •๋ณด๋ฅผ ์ œ๊ณตํ•˜๋Š”์ง€ ํŒ๋‹จํ•ฉ๋‹ˆ๋‹ค. ๊ณ ํ’ˆ์งˆ ๋‹ต๋ณ€์€ ์งˆ๋ฌธ์ž์—๊ฒŒ ์œ ์šฉํ•˜๊ณ  ๊ด€๋ จ์„ฑ ์žˆ๋Š” ์ •๋ณด๋ฅผ ์ œ๊ณตํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
In this section, we present our findings on the predictability of model losses regarding data mixtures, which boils down to functional relationships we refer to as the data mixing laws.
์ด ์„น์…˜์—์„œ๋Š” ๋ฐ์ดํ„ฐ ํ˜ผํ•ฉ์— ๊ด€ํ•œ ๋ชจ๋ธ ์†์‹ค์˜ ์˜ˆ์ธก ๊ฐ€๋Šฅ์„ฑ์— ๋Œ€ํ•œ ์šฐ๋ฆฌ์˜ ๋ฐœ๊ฒฌ์„ ์ œ์‹œํ•ฉ๋‹ˆ๋‹ค. ์ด๋Š” ์šฐ๋ฆฌ๊ฐ€ ๋ฐ์ดํ„ฐ ํ˜ผํ•ฉ ๋ฒ•์น™์ด๋ผ๊ณ  ๋ถ€๋ฅด๋Š” ๊ธฐ๋Šฅ์  ๊ด€๊ณ„๋กœ ์š”์•ฝ๋ฉ๋‹ˆ๋‹ค.
์ด ์ ˆ์—์„œ๋Š” ๋ฐ์ดํ„ฐ ํ˜ผํ•ฉ๊ณผ ๊ด€๋ จ๋œ ๋ชจ๋ธ ์†์‹ค์˜ ์˜ˆ์ธก ๊ฐ€๋Šฅ์„ฑ์— ๋Œ€ํ•œ ์—ฐ๊ตฌ ๊ฒฐ๊ณผ๋ฅผ ์ œ์‹œํ•˜๋ฉฐ, ์ด๋Š” ๋ฐ์ดํ„ฐ ํ˜ผํ•ฉ ๋ฒ•์น™์ด๋ผ๊ณ  ํ•˜๋Š” ํ•จ์ˆ˜ ๊ด€๊ณ„๋กœ ๊ท€๊ฒฐ๋œ๋‹ค.
Fig 24 provides an overview of some of the most representative LLM frameworks, and the relevant works that have contributed to the success of LLMs and helped to push the limits of LLMs.
Fig 24๋Š” ๊ฐ€์žฅ ๋Œ€ํ‘œ์ ์ธ LLM ํ”„๋ ˆ์ž„์›Œํฌ์™€ LLM์˜ ์„ฑ๊ณต์— ๊ธฐ์—ฌํ•˜๊ณ  LLM์˜ ํ•œ๊ณ„๋ฅผ ํ™•์žฅํ•˜๋Š” ๋ฐ ๋„์›€์„ ์ค€ ๊ด€๋ จ ์ž‘์—…๋“ค์˜ ๊ฐœ์š”๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.
๊ทธ๋ฆผ 24๋Š” ๊ฐ€์žฅ ๋Œ€ํ‘œ์ ์ธ LLM ํ”„๋ ˆ์ž„์›Œํฌ์™€ LLM์˜ ์„ฑ๊ณต์— ๊ธฐ์—ฌํ•˜๊ณ  LLM์˜ ํ•œ๊ณ„๋ฅผ ๋›ฐ์–ด๋„˜๋Š” ๋ฐ ๋„์›€์„ ์ค€ ๊ด€๋ จ ์—ฐ๊ตฌ์— ๋Œ€ํ•œ ๊ฐœ์š”๋ฅผ ์ œ๊ณตํ•œ๋‹ค.
This underscores the _crucial importance of pretrain data quality_: even if junk data is entirely random, it negatively impacts model's knowledge capacity even with sufficient training.
์ด๋Š” _์‚ฌ์ „ ํ•™์Šต ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ์˜ ๊ฒฐ์ •์ ์ธ ์ค‘์š”์„ฑ_์„ ๊ฐ•์กฐํ•ฉ๋‹ˆ๋‹ค: ์“ฐ๋ ˆ๊ธฐ ๋ฐ์ดํ„ฐ๊ฐ€ ์™„์ „ํžˆ ๋ฌด์ž‘์œ„์ผ์ง€๋ผ๋„, ์ถฉ๋ถ„ํ•œ ํ›ˆ๋ จ์ด ์ด๋ฃจ์–ด์กŒ๋”๋ผ๋„ ๋ชจ๋ธ์˜ ์ง€์‹ ์šฉ๋Ÿ‰์— ๋ถ€์ •์ ์ธ ์˜ํ–ฅ์„ ๋ฏธ์นฉ๋‹ˆ๋‹ค.
์ด๋Š” _์‚ฌ์ „ ํ•™์Šต ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ์˜ ์ค‘์š”์„ฑ_์„ ๊ฐ•์กฐํ•ฉ๋‹ˆ๋‹ค. ์ •ํฌ ๋ฐ์ดํ„ฐ๊ฐ€ ์™„์ „ํžˆ ๋ฌด์ž‘์œ„๋ผ ํ•˜๋”๋ผ๋„ ์ถฉ๋ถ„ํ•œ ํ›ˆ๋ จ์„ ํ†ตํ•ด์„œ๋„ ๋ชจ๋ธ์˜ ์ง€์‹ ๋Šฅ๋ ฅ์— ๋ถ€์ •์ ์ธ ์˜ํ–ฅ์„ ๋ฏธ์นฉ๋‹ˆ๋‹ค.
To add these fractions, we need to find a common denominator, which is the least common multiple (LCM) of the three denominators (-4, -2, and -3). The LCM of these denominators is -12.
์ด ๋ถ„์ˆ˜๋“ค์„ ๋”ํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์„ธ ๋ถ„๋ชจ(-4, -2, -3)์˜ ์ตœ์†Œ ๊ณต๋ฐฐ์ˆ˜(LCM)๋ฅผ ์ฐพ์•„์•ผ ํ•ฉ๋‹ˆ๋‹ค. ์ด ๋ถ„๋ชจ๋“ค์˜ LCM์€ -12์ž…๋‹ˆ๋‹ค.
์ด ๋ถ„์ˆ˜๋ฅผ ๋”ํ•˜๋ ค๋ฉด ์„ธ ๋ถ„๋ชจ(-4, -2, -3)์˜ ์ตœ์†Œ๊ณต๋ฐฐ์ˆ˜(LCM)์ธ ๊ณตํ†ต ๋ถ„๋ชจ๋ฅผ ์ฐพ์•„์•ผ ํ•ฉ๋‹ˆ๋‹ค. ์ด ๋ถ„๋ชจ์˜ LCM์€ -12์ž…๋‹ˆ๋‹ค.
Generate a conversation of around four turns between PersonA and PersonU where PersonU needs information from PersonA that is in the passage. Use only information found in the passage.
PersonA์™€ PersonU ์‚ฌ์ด์— ์•ฝ ๋„ค ๋ฒˆ์˜ ๋Œ€ํ™”๋ฅผ ์ƒ์„ฑํ•˜์„ธ์š”. ์ด๋•Œ PersonU๋Š” PersonA๋กœ๋ถ€ํ„ฐ ๋ณธ๋ฌธ์— ์žˆ๋Š” ์ •๋ณด๋ฅผ ํ•„์š”๋กœ ํ•ฉ๋‹ˆ๋‹ค. ๋ณธ๋ฌธ์—์„œ ์ฐพ์„ ์ˆ˜ ์žˆ๋Š” ์ •๋ณด๋งŒ ์‚ฌ์šฉํ•˜์„ธ์š”.
PersonA์™€ PersonU๊ฐ€ 4๋ฒˆ ์ •๋„ ๋Œ€ํ™”๋ฅผ ๋‚˜๋ˆ„๊ณ , PersonU๊ฐ€ ์ง€๋ฌธ์— ์žˆ๋Š” PersonA์˜ ์ •๋ณด๊ฐ€ ํ•„์š”ํ•œ ๊ฒฝ์šฐ, ์ง€๋ฌธ์— ์žˆ๋Š” ์ •๋ณด๋งŒ ์‚ฌ์šฉํ•˜์—ฌ ๋Œ€ํ™”๋ฅผ ์ž‘์„ฑํ•ฉ๋‹ˆ๋‹ค.
Figure 22: Examples on emoji. Both models understand emojis and structured data well, and accurately control the number of items implicitly. The DPO model tends to generate more words.
Figure 22: ์ด๋ชจ์ง€์— ๋Œ€ํ•œ ์˜ˆ์‹œ. ๋‘ ๋ชจ๋ธ ๋ชจ๋‘ ์ด๋ชจ์ง€์™€ ๊ตฌ์กฐํ™”๋œ ๋ฐ์ดํ„ฐ๋ฅผ ์ž˜ ์ดํ•ดํ•˜๋ฉฐ, ํ•ญ๋ชฉ์˜ ์ˆ˜๋ฅผ ์•”๋ฌต์ ์œผ๋กœ ์ •ํ™•ํ•˜๊ฒŒ ์ œ์–ดํ•ฉ๋‹ˆ๋‹ค. DPO ๋ชจ๋ธ์€ ๋” ๋งŽ์€ ๋‹จ์–ด๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๊ฒฝํ–ฅ์ด ์žˆ์Šต๋‹ˆ๋‹ค.
๊ทธ๋ฆผ 22: ์ด๋ชจํ‹ฐ์ฝ˜์˜ ์˜ˆ. ๋‘ ๋ชจ๋ธ ๋ชจ๋‘ ์ด๋ชจํ‹ฐ์ฝ˜๊ณผ ๊ตฌ์กฐํ™”๋œ ๋ฐ์ดํ„ฐ๋ฅผ ์ž˜ ์ดํ•ดํ•˜๊ณ  ํ•ญ๋ชฉ ์ˆ˜๋ฅผ ์•”์‹œ์ ์œผ๋กœ ์ •ํ™•ํ•˜๊ฒŒ ์ œ์–ดํ•ฉ๋‹ˆ๋‹ค. DPO ๋ชจ๋ธ์€ ๋” ๋งŽ์€ ๋‹จ์–ด๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๊ฒฝํ–ฅ์ด ์žˆ์Šต๋‹ˆ๋‹ค.
Footnote 10: Grokking refers that โ€œa pattern in the data, improving generalization performance from random chance level to perfect generalizationโ€, quoted from the original paper [73].
Footnote 10: Grokking์€ "๋ฐ์ดํ„ฐ์—์„œ ํŒจํ„ด์„ ๋ฐœ๊ฒฌํ•˜์—ฌ ์ผ๋ฐ˜ํ™” ์„ฑ๋Šฅ์„ ๋ฌด์ž‘์œ„ ์ˆ˜์ค€์—์„œ ์™„๋ฒฝํ•œ ์ผ๋ฐ˜ํ™”๋กœ ํ–ฅ์ƒ์‹œํ‚ค๋Š” ๊ฒƒ"์„ ์˜๋ฏธํ•˜๋ฉฐ, ์ด๋Š” ์› ๋…ผ๋ฌธ [73]์—์„œ ์ธ์šฉ๋œ ๊ฒƒ์ž…๋‹ˆ๋‹ค.
๊ฐ์ฃผ 10: ๊ทธ๋กœํ‚น์€ ๋ฐ์ดํ„ฐ์˜ ํŒจํ„ด์„ ์˜๋ฏธํ•˜๋ฉฐ, ์› ๋…ผ๋ฌธ[73]์—์„œ ์ธ์šฉํ•œ ๊ฒƒ์œผ๋กœ "๋ฌด์ž‘์œ„ ์šฐ์—ฐ ์ˆ˜์ค€์—์„œ ์™„๋ฒฝํ•œ ์ผ๋ฐ˜ํ™” ์„ฑ๋Šฅ๊นŒ์ง€ ์ผ๋ฐ˜ํ™” ์„ฑ๋Šฅ์„ ํ–ฅ์ƒ์‹œํ‚จ๋‹ค"๋Š” ์˜๋ฏธ์ž…๋‹ˆ๋‹ค.
_For example, if \(T_{0}\gg T\) (e.g \(T_{0}\geq CT\log T\)) and \(n\) is constant (e.g \(n=25\)), then model collapse will not occur if we learn on the \(n\)th generation of AI data.
์˜ˆ๋ฅผ ๋“ค์–ด, \(T_{0}\gg T\) (์˜ˆ: \(T_{0}\geq CT\log T\))์ด๊ณ  \(n\)์ด ์ƒ์ˆ˜์ผ ๋•Œ (์˜ˆ: \(n=25\)), AI ๋ฐ์ดํ„ฐ์˜ \(n\)๋ฒˆ์งธ ์„ธ๋Œ€์—์„œ ํ•™์Šตํ•˜๋ฉด ๋ชจ๋ธ ๋ถ•๊ดด๊ฐ€ ๋ฐœ์ƒํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
_์˜ˆ๋ฅผ ๋“ค์–ด, \(T_{0}\gg T\) (์˜ˆ: \(T_{0}\geq CT\log T\))์ด๊ณ  \(n\)์ด ์ƒ์ˆ˜์ธ ๊ฒฝ์šฐ(์˜ˆ: \(n=25\)) AI ๋ฐ์ดํ„ฐ์˜ \(n\)๋ฒˆ์งธ ์„ธ๋Œ€์—์„œ ํ•™์Šตํ•˜๋ฉด ๋ชจ๋ธ ๋ถ•๊ดด๊ฐ€ ๋ฐœ์ƒํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
Generate a conversation of around two turns between PersonA and PersonU where PersonU needs information from PersonA that is in the passage. Use only information found in the passage.
PersonA์™€ PersonU ์‚ฌ์ด์˜ ๋Œ€ํ™”๋ฅผ ์•ฝ ๋‘ ๋ฒˆ ์ฃผ๊ณ ๋ฐ›๋Š” ํ˜•ํƒœ๋กœ ์ƒ์„ฑํ•˜์„ธ์š”. ์ด๋•Œ PersonU๋Š” PersonA์—๊ฒŒ์„œ ๋ณธ๋ฌธ์— ์žˆ๋Š” ์ •๋ณด๋ฅผ ํ•„์š”๋กœ ํ•ฉ๋‹ˆ๋‹ค. ๋ณธ๋ฌธ์—์„œ ์ฐพ์„ ์ˆ˜ ์žˆ๋Š” ์ •๋ณด๋งŒ ์‚ฌ์šฉํ•˜์„ธ์š”.
PersonA์™€ PersonU ์‚ฌ์ด์˜ ์•ฝ ๋‘ ๋ฒˆ์˜ ๋Œ€ํ™”๋ฅผ ์ƒ์„ฑํ•˜๊ณ , PersonU๊ฐ€ ๊ตฌ์ ˆ์— ์žˆ๋Š” PersonA์˜ ์ •๋ณด๊ฐ€ ํ•„์š”ํ•œ ๊ฒฝ์šฐ์ž…๋‹ˆ๋‹ค. ์ง€๋ฌธ์—์„œ ์ฐพ์€ ์ •๋ณด๋งŒ ์‚ฌ์šฉํ•˜์„ธ์š”.
- \(\{\)(16-3-4=9\)\(\}\) So our daily profit is 9 times $2, or \(\{\)(9times\()$2=\$18\(\}\)) Hence her profits at market on a given day are \(\{\)$18\(\}\). Answer: \(\{\)$18\(\}\).
- \(\{\)(16-3-4=9\)\(\}\) ๋”ฐ๋ผ์„œ ์šฐ๋ฆฌ์˜ ์ผ์ผ ์ˆ˜์ต์€ 9์— $2๋ฅผ ๊ณฑํ•œ ๊ฐ’, ์ฆ‰ \(\{\)(9times\()$2=\$18\(\}\))์ž…๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ๊ทธ๋…€์˜ ํŠน์ • ๋‚  ์‹œ์žฅ์—์„œ์˜ ์ˆ˜์ต์€ \(\{\)$18\(\}\)์ž…๋‹ˆ๋‹ค. ๋‹ต: \(\{\)$18\(\}\).
- \(\{\)(16-3-4=9\)\(\}\) ๋”ฐ๋ผ์„œ ์ผ์ผ ์ˆ˜์ต์€ 2๋‹ฌ๋Ÿฌ์˜ 9๋ฐฐ, ์ฆ‰ \(\{\)(9times\()$2=\$18\(\}\)) ๋”ฐ๋ผ์„œ ์ฃผ์–ด์ง„ ๋‚ ์งœ์— ๊ทธ๋…€์˜ ์‹œ์žฅ ์ˆ˜์ต์€ \(\{\)$18\(\}\). Answer: \(\{\)$18\(\}\).
_Remark 1.5_.: Since int8 is 8bit, LLMs can exceed 1/4 of the theoretical limit for storing knowledge; thus knowledge must be very compactly stored inside the model across all layers.
_Remark 1.5_.: int8์ด 8๋น„ํŠธ์ด๊ธฐ ๋•Œ๋ฌธ์—, LLM์€ ์ง€์‹์„ ์ €์žฅํ•˜๋Š” ์ด๋ก ์  ํ•œ๊ณ„์˜ 1/4์„ ์ดˆ๊ณผํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ง€์‹์€ ๋ชจ๋“  ๋ ˆ์ด์–ด์— ๊ฑธ์ณ ๋ชจ๋ธ ๋‚ด๋ถ€์— ๋งค์šฐ ์••์ถ•๋œ ํ˜•ํƒœ๋กœ ์ €์žฅ๋˜์–ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
_Remark 1.5_.: int8์€ 8๋น„ํŠธ์ด๋ฏ€๋กœ LLM์€ ์ด๋ก ์ƒ ์ง€์‹ ์ €์žฅ ํ•œ๊ณ„์˜ 1/4์„ ์ดˆ๊ณผํ•  ์ˆ˜ ์žˆ์œผ๋ฏ€๋กœ ์ง€์‹์€ ๋ชจ๋“  ๋ ˆ์ด์–ด์— ๊ฑธ์ณ ๋ชจ๋ธ ๋‚ด๋ถ€์— ๋งค์šฐ ์••์ถ•๋˜์–ด ์ €์žฅ๋˜์–ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
When we assume that the AI-generator itself was trained on \(T_{0}\) samples, we get a similar loss of scaling stemming from the tail cutting from finite sampling bias (Equation (6)):
AI ์ƒ์„ฑ๊ธฐ๊ฐ€ \(T_{0}\) ์ƒ˜ํ”Œ๋กœ ํ›ˆ๋ จ๋˜์—ˆ๋‹ค๊ณ  ๊ฐ€์ •ํ•˜๋ฉด, ์œ ํ•œ ์ƒ˜ํ”Œ๋ง ํŽธํ–ฅ์œผ๋กœ ์ธํ•œ ๊ผฌ๋ฆฌ ์ ˆ๋‹จ์—์„œ ๋น„๋กฏ๋œ ์œ ์‚ฌํ•œ ์Šค์ผ€์ผ๋ง ์†์‹ค์„ ์–ป๊ฒŒ ๋ฉ๋‹ˆ๋‹ค (Equation (6)):
AI ์ƒ์„ฑ๊ธฐ ์ž์ฒด๊ฐ€ \(T_{0}\) ์ƒ˜ํ”Œ์— ๋Œ€ํ•ด ํ•™์Šต๋œ ๊ฒƒ์œผ๋กœ ๊ฐ€์ •ํ•˜๋ฉด, ์œ ํ•œ ์ƒ˜ํ”Œ๋ง ํŽธํ–ฅ(๋ฐฉ์ •์‹ (6))์œผ๋กœ ์ธํ•œ ๊ผฌ๋ฆฌ ์ ˆ๋‹จ์—์„œ ๋น„๋กฏ๋œ ์œ ์‚ฌํ•œ ์Šค์ผ€์ผ๋ง ์†์‹ค์„ ์–ป๊ฒŒ ๋ฉ๋‹ˆ๋‹ค:
**Comparison with Self-Sampling.** RFT collects the reasoning paths sampled by the SFT model itself and filters out those paths that contain incorrect answers. Our experiments in Tab.
**Self-Sampling๊ณผ์˜ ๋น„๊ต.** RFT๋Š” SFT ๋ชจ๋ธ ์ž์ฒด์—์„œ ์ƒ˜ํ”Œ๋งํ•œ ์ถ”๋ก  ๊ฒฝ๋กœ๋ฅผ ์ˆ˜์ง‘ํ•˜๊ณ , ์ž˜๋ชป๋œ ๋‹ต๋ณ€์„ ํฌํ•จํ•˜๋Š” ๊ฒฝ๋กœ๋ฅผ ํ•„ํ„ฐ๋งํ•ฉ๋‹ˆ๋‹ค. ์šฐ๋ฆฌ์˜ ์‹คํ—˜์€ Tab.
**์ž์ฒด ์ƒ˜ํ”Œ๋ง๊ณผ ๋น„๊ต**. RFT๋Š” SFT ๋ชจ๋ธ ์ž์ฒด์—์„œ ์ƒ˜ํ”Œ๋งํ•œ ์ถ”๋ก  ๊ฒฝ๋กœ๋ฅผ ์ˆ˜์ง‘ํ•˜๊ณ  ์˜ค๋‹ต์ด ํฌํ•จ๋œ ๊ฒฝ๋กœ๋ฅผ ํ•„ํ„ฐ๋งํ•ฉ๋‹ˆ๋‹ค. Tab.
Table 3 showcases examples of interactions with our _WizardCoder_. The examples demonstrate that our model consistently generates accurate responses accompanied by clear explanations.
Table 3์€ _WizardCoder_์™€์˜ ์ƒํ˜ธ์ž‘์šฉ ์˜ˆ์‹œ๋ฅผ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค. ์ด ์˜ˆ์‹œ๋“ค์€ ์šฐ๋ฆฌ ๋ชจ๋ธ์ด ๋ช…ํ™•ํ•œ ์„ค๋ช…๊ณผ ํ•จ๊ป˜ ์ผ๊ด€๋˜๊ฒŒ ์ •ํ™•ํ•œ ์‘๋‹ต์„ ์ƒ์„ฑํ•จ์„ ์ž…์ฆํ•ฉ๋‹ˆ๋‹ค.
ํ‘œ 3์€ _WizardCoder_์™€์˜ ์ƒํ˜ธ ์ž‘์šฉ์˜ ์˜ˆ๋ฅผ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค. ์ด ์˜ˆ๋Š” ์šฐ๋ฆฌ ๋ชจ๋ธ์ด ๋ช…ํ™•ํ•œ ์„ค๋ช…๊ณผ ํ•จ๊ป˜ ์ผ๊ด€๋˜๊ฒŒ ์ •ํ™•ํ•œ ์‘๋‹ต์„ ์ƒ์„ฑํ•œ๋‹ค๋Š” ๊ฒƒ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
Therefore, although 1000 exposures may be necessary for a model to reach its maximum storage capacity, training with just 100 exposures results in a capacity loss of no more than 2x.
๋”ฐ๋ผ์„œ ๋ชจ๋ธ์ด ์ตœ๋Œ€ ์ €์žฅ ์šฉ๋Ÿ‰์— ๋„๋‹ฌํ•˜๊ธฐ ์œ„ํ•ด 1000๋ฒˆ์˜ ๋…ธ์ถœ์ด ํ•„์š”ํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, ๋‹จ 100๋ฒˆ์˜ ๋…ธ์ถœ๋กœ ํ›ˆ๋ จํ•ด๋„ ์šฉ๋Ÿ‰ ์†์‹ค์€ ์ตœ๋Œ€ 2๋ฐฐ๋ฅผ ๋„˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
๋”ฐ๋ผ์„œ ๋ชจ๋ธ์ด ์ตœ๋Œ€ ์ €์žฅ ์šฉ๋Ÿ‰์— ๋„๋‹ฌํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” 1000ํšŒ ๋…ธ์ถœ์ด ํ•„์š”ํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, 100ํšŒ ๋…ธ์ถœ๋กœ๋งŒ ํ•™์Šตํ•ด๋„ ์šฉ๋Ÿ‰ ์†์‹ค์€ 2๋ฐฐ ์ดํ•˜์ž…๋‹ˆ๋‹ค.
Since Adrien's total salary was 30 percent higher than Lylah's, and Adrien earned $40000 four years ago, we can find Lylah's salary by subtracting 30% of $40000 from Adrien's salary.
Adrien์˜ ์ด ๊ธ‰์—ฌ๊ฐ€ Lylah์˜ ๊ธ‰์—ฌ๋ณด๋‹ค 30% ๋” ๋†’์•˜๊ณ , Adrien์ด 4๋…„ ์ „์— $40,000๋ฅผ ๋ฒŒ์—ˆ๊ธฐ ๋•Œ๋ฌธ์—, Adrien์˜ ๊ธ‰์—ฌ์—์„œ $40,000์˜ 30%๋ฅผ ๋นผ๋ฉด Lylah์˜ ๊ธ‰์—ฌ๋ฅผ ์•Œ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
์•„๋“œ๋ฆฌ์•ˆ์˜ ์ด ๊ธ‰์—ฌ๊ฐ€ ๋ผ์ผ๋ผ์˜ ๊ธ‰์—ฌ๋ณด๋‹ค 30% ๋” ๋†’์•˜๊ณ , ์•„๋“œ๋ฆฌ์•ˆ์ด 4๋…„ ์ „์— 40000๋‹ฌ๋Ÿฌ๋ฅผ ๋ฒŒ์—ˆ์œผ๋ฏ€๋กœ, ์•„๋“œ๋ฆฌ์•ˆ์˜ ๊ธ‰์—ฌ์—์„œ 40000๋‹ฌ๋Ÿฌ์˜ 30%๋ฅผ ๋นผ๋ฉด ๋ผ์ผ๋ผ์˜ ๊ธ‰์—ฌ๋ฅผ ๊ตฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
The comprehensive results of WMT'23 are presented in Table 14. Similar to its performance in WMT'21 and WMT'22, ALMA-13B-R performs best on average among the SoTA translation models.
WMT'23์˜ ์ข…ํ•ฉ ๊ฒฐ๊ณผ๋Š” Table 14์— ์ œ์‹œ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. WMT'21 ๋ฐ WMT'22์—์„œ์˜ ์„ฑ๋Šฅ๊ณผ ์œ ์‚ฌํ•˜๊ฒŒ, ALMA-13B-R์€ SoTA ๋ฒˆ์—ญ ๋ชจ๋ธ ์ค‘ ํ‰๊ท ์ ์œผ๋กœ ๊ฐ€์žฅ ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์„ ๋ณด์ž…๋‹ˆ๋‹ค.
WMT'23์˜ ์ข…ํ•ฉ ๊ฒฐ๊ณผ๋Š” ํ‘œ 14์— ์ œ์‹œ๋˜์–ด ์žˆ๋‹ค. WMT'21๊ณผ WMT'22์—์„œ์˜ ์„ฑ๋Šฅ๊ณผ ์œ ์‚ฌํ•˜๊ฒŒ, ALMA-13B-R์€ SoTA ๋ฒˆ์—ญ ๋ชจ๋ธ ์ค‘ ํ‰๊ท ์ ์œผ๋กœ ๊ฐ€์žฅ ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ์„ ๋ณด์ธ๋‹ค.
logits produced by the original model. This method ensures the efficient scaling of the model by optimizing its architecture based on the internal processing dynamics of its layers.
์›๋ณธ ๋ชจ๋ธ์— ์˜ํ•ด ์ƒ์„ฑ๋œ ๋กœ์ง“. ์ด ๋ฐฉ๋ฒ•์€ ๋ ˆ์ด์–ด์˜ ๋‚ด๋ถ€ ์ฒ˜๋ฆฌ ๋™์—ญํ•™์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์•„ํ‚คํ…์ฒ˜๋ฅผ ์ตœ์ ํ™”ํ•˜์—ฌ ๋ชจ๋ธ์˜ ํšจ์œจ์ ์ธ ํ™•์žฅ์„ ๋ณด์žฅํ•ฉ๋‹ˆ๋‹ค.
์›๋ž˜ ๋ชจ๋ธ์—์„œ ์ƒ์„ฑ๋œ ๋กœ์ง“์„ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค. ์ด ๋ฐฉ๋ฒ•์€ ๋ ˆ์ด์–ด์˜ ๋‚ด๋ถ€ ์ฒ˜๋ฆฌ ๋™์—ญํ•™์— ๋”ฐ๋ผ ๋ชจ๋ธ ์•„ํ‚คํ…์ฒ˜๋ฅผ ์ตœ์ ํ™”ํ•˜์—ฌ ๋ชจ๋ธ์˜ ํšจ์œจ์ ์ธ ํ™•์žฅ์„ ๋ณด์žฅํ•ฉ๋‹ˆ๋‹ค.
_(B) If \(C>1\), then \(E_{test}\asymp T^{-c}+k^{-\alpha}\) (which recovers the result of Theorem 2.1), and so "buying" the \(N\)th tail of the real data distribution is worthless._
(B) ๋งŒ์•ฝ \(C>1\)์ด๋ผ๋ฉด, \(E_{test}\asymp T^{-c}+k^{-\alpha}\) (์ด๋Š” Theorem 2.1์˜ ๊ฒฐ๊ณผ๋ฅผ ํšŒ๋ณตํ•ฉ๋‹ˆ๋‹ค), ๋”ฐ๋ผ์„œ ์‹ค์ œ ๋ฐ์ดํ„ฐ ๋ถ„ํฌ์˜ \(N\)๋ฒˆ์งธ ๊ผฌ๋ฆฌ๋ฅผ "๊ตฌ๋งค"ํ•˜๋Š” ๊ฒƒ์€ ๋ฌด์˜๋ฏธํ•ฉ๋‹ˆ๋‹ค.
_(B) C>1\)์ด๋ฉด \(E_{test}\asymp T^{-c}+k^{-\alpha}\)(์ •๋ฆฌ 2.1์˜ ๊ฒฐ๊ณผ๋ฅผ ํšŒ๋ณต)๊ฐ€ ๋˜๋ฏ€๋กœ ์‹ค์ œ ๋ฐ์ดํ„ฐ ๋ถ„ํฌ์˜ \(N\)๋ฒˆ์งธ ๊ผฌ๋ฆฌ๋ฅผ "๊ตฌ๋งค"ํ•˜๋Š” ๊ฒƒ์€ ๊ฐ€์น˜๊ฐ€ ์—†์Šต๋‹ˆ๋‹ค.
Table 2 shows that the efficacy of our method varies with the difficulty of the task. In this section, we aim to understand the underlying properties through controlled experiments.
Table 2๋Š” ์šฐ๋ฆฌ์˜ ๋ฐฉ๋ฒ•์˜ ํšจ๋Šฅ์ด ์ž‘์—…์˜ ๋‚œ์ด๋„์— ๋”ฐ๋ผ ๋‹ฌ๋ผ์ง„๋‹ค๋Š” ๊ฒƒ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค. ์ด ์„น์…˜์—์„œ๋Š” ํ†ต์ œ๋œ ์‹คํ—˜์„ ํ†ตํ•ด ๊ธฐ๋ณธ์ ์ธ ํŠน์„ฑ์„ ์ดํ•ดํ•˜๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•ฉ๋‹ˆ๋‹ค.
ํ‘œ 2๋Š” ์ž‘์—…์˜ ๋‚œ์ด๋„์— ๋”ฐ๋ผ ์šฐ๋ฆฌ ๋ฐฉ๋ฒ•์˜ ํšจ๋Šฅ์ด ๋‹ฌ๋ผ์ง์„ ๋ณด์—ฌ์ค€๋‹ค. ์ด ์„น์…˜์—์„œ๋Š” ํ†ต์ œ๋œ ์‹คํ—˜์„ ํ†ตํ•ด ๊ทผ๋ณธ์ ์ธ ์†์„ฑ์„ ์ดํ•ดํ•˜๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•œ๋‹ค.
Admittedly, how data mixtures affect model training is far from fully understood. Our study makes preliminary attempts at a quantitative framework while leaving several limitations.
์†”์งํžˆ ๋งํ•ด์„œ, ๋ฐ์ดํ„ฐ ํ˜ผํ•ฉ์ด ๋ชจ๋ธ ํ›ˆ๋ จ์— ๋ฏธ์น˜๋Š” ์˜ํ–ฅ์€ ์•„์ง ์™„์ „ํžˆ ์ดํ•ด๋˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ์˜ ์—ฐ๊ตฌ๋Š” ๋ช‡ ๊ฐ€์ง€ ํ•œ๊ณ„๋ฅผ ๋‚จ๊ธฐ๋ฉด์„œ ์ •๋Ÿ‰์  ํ”„๋ ˆ์ž„์›Œํฌ์— ๋Œ€ํ•œ ์ดˆ๊ธฐ ์‹œ๋„๋ฅผ ํ•ฉ๋‹ˆ๋‹ค.
์ธ์ •ํ•  ๋•Œ, ๋ฐ์ดํ„ฐ ํ˜ผํ•ฉ์ด ๋ชจ๋ธ ํ•™์Šต์— ์–ด๋–ค ์˜ํ–ฅ์„ ๋ฏธ์น˜๋Š”์ง€๋Š” ์™„์ „ํžˆ ์ดํ•ด๋˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ์˜ ์—ฐ๊ตฌ๋Š” ๋ช‡ ๊ฐ€์ง€ ํ•œ๊ณ„๋ฅผ ๋‚จ๊ธฐ๋ฉด์„œ ์ •๋Ÿ‰์  ํ”„๋ ˆ์ž„์›Œํฌ์— ๋Œ€ํ•œ ์˜ˆ๋น„์ ์ธ ์‹œ๋„๋ฅผ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
Table 3 compares the performance of our VLM with the baselines. Please note that the Japanese Stable VLM cannot be evaluated on JA-VG-VQA-500 because it was trained on this dataset.
Table 3์€ ์šฐ๋ฆฌ์˜ VLM๊ณผ ๊ธฐ์ค€ ๋ชจ๋ธ๋“ค์˜ ์„ฑ๋Šฅ์„ ๋น„๊ตํ•ฉ๋‹ˆ๋‹ค. ์ผ๋ณธ์–ด Stable VLM์€ ์ด ๋ฐ์ดํ„ฐ์…‹์œผ๋กœ ํ›ˆ๋ จ๋˜์—ˆ๊ธฐ ๋•Œ๋ฌธ์— JA-VG-VQA-500์—์„œ ํ‰๊ฐ€ํ•  ์ˆ˜ ์—†์Œ์„ ์œ ์˜ํ•ด ์ฃผ์„ธ์š”.
ํ‘œ 3์€ ๋ฒ ์ด์Šค๋ผ์ธ๊ณผ ์šฐ๋ฆฌ VLM์˜ ์„ฑ๋Šฅ์„ ๋น„๊ตํ•œ ๊ฒƒ์ด๋‹ค. ์ผ๋ณธ์–ด Stable VLM์€ ์ด ๋ฐ์ดํ„ฐ ์„ธํŠธ์—์„œ ํ•™์Šต๋˜์—ˆ๊ธฐ ๋•Œ๋ฌธ์— JA-VG-VQA-500์—์„œ ํ‰๊ฐ€ํ•  ์ˆ˜ ์—†๋‹ค๋Š” ์ ์— ์œ ์˜ํ•˜์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.
Footnote 25: Due to the limit of computational resources, we cannot conduct large-scale experiments on larger LLaMA variants right now, which would be scheduled in a future version.
Footnote 25: ๊ณ„์‚ฐ ์ž์›์˜ ํ•œ๊ณ„๋กœ ์ธํ•ด ํ˜„์žฌ ๋” ํฐ LLaMA ๋ณ€์ข…์— ๋Œ€ํ•œ ๋Œ€๊ทœ๋ชจ ์‹คํ—˜์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์—†์œผ๋ฉฐ, ์ด๋Š” ํ–ฅํ›„ ๋ฒ„์ „์—์„œ ์˜ˆ์ •๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค.
๊ฐ์ฃผ 25: ๊ณ„์‚ฐ ๋ฆฌ์†Œ์Šค์˜ ํ•œ๊ณ„๋กœ ์ธํ•ด ํ˜„์žฌ๋กœ์„œ๋Š” ๋” ํฐ ๊ทœ๋ชจ์˜ LLaMA ๋ณ€ํ˜•์— ๋Œ€ํ•œ ๋Œ€๊ทœ๋ชจ ์‹คํ—˜์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์—†์œผ๋ฉฐ, ์ด๋Š” ํ–ฅํ›„ ๋ฒ„์ „์—์„œ ์˜ˆ์ •๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
entropy, suggesting that unlabeled task data contains new knowledge but is less diverse. For ETA-DACP-ppl and ETA-DACP-ent, the samples are uniform across the other two dimensions.
์—”ํŠธ๋กœํ”ผ, ์ด๋Š” ๋ผ๋ฒจ์ด ์—†๋Š” ์ž‘์—… ๋ฐ์ดํ„ฐ๊ฐ€ ์ƒˆ๋กœ์šด ์ง€์‹์„ ํฌํ•จํ•˜๊ณ  ์žˆ์ง€๋งŒ ๋‹ค์–‘์„ฑ์€ ์ ๋‹ค๋Š” ๊ฒƒ์„ ์‹œ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ETA-DACP-ppl๊ณผ ETA-DACP-ent์˜ ๊ฒฝ์šฐ, ์ƒ˜ํ”Œ์€ ๋‹ค๋ฅธ ๋‘ ์ฐจ์›์—์„œ ๊ท ์ผํ•ฉ๋‹ˆ๋‹ค.
์—”ํŠธ๋กœํ”ผ๋Š” ๋ ˆ์ด๋ธ”์ด ์—†๋Š” ์ž‘์—… ๋ฐ์ดํ„ฐ์— ์ƒˆ๋กœ์šด ์ง€์‹์ด ํฌํ•จ๋˜์–ด ์žˆ์ง€๋งŒ ๋‹ค์–‘์„ฑ์ด ๋–จ์–ด์ง„๋‹ค๋Š” ๊ฒƒ์„ ์‹œ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ETA-DACP-ppl๊ณผ ETA-DACP-ent์˜ ๊ฒฝ์šฐ, ๋‹ค๋ฅธ ๋‘ ์ฐจ์›์— ๊ฑธ์ณ ์ƒ˜ํ”Œ์ด ๊ท ์ผํ•ฉ๋‹ˆ๋‹ค.
5. Now, we have another instance of \(f\) and \(g\). Since \(f\) is odd and \(g\) is even, their product is an odd function. So, we can rewrite the expression as \(-[f(g(f(x)))]\).
์ด์ œ \(f\)์™€ \(g\)์˜ ๋˜ ๋‹ค๋ฅธ ์˜ˆ๋ฅผ ์‚ดํŽด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. \(f\)๋Š” ํ™€์ˆ˜ ํ•จ์ˆ˜์ด๊ณ  \(g\)๋Š” ์ง์ˆ˜ ํ•จ์ˆ˜์ด๋ฏ€๋กœ, ์ด๋“ค์˜ ๊ณฑ์€ ํ™€์ˆ˜ ํ•จ์ˆ˜๊ฐ€ ๋ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ด ํ‘œํ˜„์„ \(-[f(g(f(x)))]\)๋กœ ๋‹ค์‹œ ์“ธ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
5. ์ด์ œ \(f\)์™€ \(g\)์˜ ๋˜ ๋‹ค๋ฅธ ์ธ์Šคํ„ด์Šค๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค. f\)๋Š” ํ™€์ˆ˜์ด๊ณ  \(g\)๋Š” ์ง์ˆ˜์ด๋ฏ€๋กœ ์ด๋“ค์˜ ๊ณฑ์€ ํ™€์ˆ˜ ํ•จ์ˆ˜์ž…๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์‹์„ \(-[f(g(f(x)))]\)๋กœ ๋‹ค์‹œ ์“ธ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
**Parameter 5** (Figure 2, 12, 13).: For GPT2 models on the \(\mathsf{bioD}\) dataset, we focus on the 1000-exposure case, with \(wd=0.01\), \(lr=0.0005\), and a batch size of 192.
**Parameter 5** (Figure 2, 12, 13).: \(\mathsf{bioD}\) ๋ฐ์ดํ„ฐ์…‹์—์„œ GPT2 ๋ชจ๋ธ์˜ ๊ฒฝ์šฐ, ์šฐ๋ฆฌ๋Š” \(wd=0.01\), \(lr=0.0005\), ๋ฐฐ์น˜ ํฌ๊ธฐ 192๋กœ 1000๋ฒˆ ๋…ธ์ถœ๋˜๋Š” ๊ฒฝ์šฐ์— ์ค‘์ ์„ ๋‘ก๋‹ˆ๋‹ค.
**๋งค๊ฐœ๋ณ€์ˆ˜ 5** (๊ทธ๋ฆผ 2, 12, 13).: \(\mathsf{bioD}\) ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ GPT2 ๋ชจ๋ธ์˜ ๊ฒฝ์šฐ, 1000 ๋…ธ์ถœ ์‚ฌ๋ก€์— ์ดˆ์ ์„ ๋งž์ถ”๊ณ  \(wd=0.01\), \(lr=0.0005\), ๋ฐฐ์น˜ ํฌ๊ธฐ๋Š” 192์ž…๋‹ˆ๋‹ค.
While LLMs are evaluated on a standard set of benchmarks designed to gauge their general understanding and reasoning capabilities, the benchmarks for agent evaluation vary greatly.
LLMs๋Š” ์ผ๋ฐ˜์ ์ธ ์ดํ•ด์™€ ์ถ”๋ก  ๋Šฅ๋ ฅ์„ ์ธก์ •ํ•˜๊ธฐ ์œ„ํ•ด ์„ค๊ณ„๋œ ํ‘œ์ค€ ๋ฒค์น˜๋งˆํฌ ์„ธํŠธ๋กœ ํ‰๊ฐ€๋˜๋Š” ๋ฐ˜๋ฉด, ์—์ด์ „ํŠธ ํ‰๊ฐ€๋ฅผ ์œ„ํ•œ ๋ฒค์น˜๋งˆํฌ๋Š” ํฌ๊ฒŒ ๋‹ค๋ฆ…๋‹ˆ๋‹ค.
LLM์€ ์ผ๋ฐ˜์ ์ธ ์ดํ•ด๋ ฅ๊ณผ ์ถ”๋ก  ๋Šฅ๋ ฅ์„ ์ธก์ •ํ•˜๊ธฐ ์œ„ํ•ด ๊ณ ์•ˆ๋œ ํ‘œ์ค€ ๋ฒค์น˜๋งˆํฌ ์„ธํŠธ์— ๋Œ€ํ•ด ํ‰๊ฐ€๋˜๋Š” ๋ฐ˜๋ฉด, ์—์ด์ „ํŠธ ํ‰๊ฐ€๋ฅผ ์œ„ํ•œ ๋ฒค์น˜๋งˆํฌ๋Š” ๋งค์šฐ ๋‹ค์–‘ํ•ฉ๋‹ˆ๋‹ค.
* [label=*] * Diverse data (rewriting the same data multiple times) does not hurt -- and may sometimes improve -- the model's capacity!Let's highlight the significance of Result 2.
* ๋‹ค์–‘ํ•œ ๋ฐ์ดํ„ฐ(๋™์ผํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์—ฌ๋Ÿฌ ๋ฒˆ ๋‹ค์‹œ ์ž‘์„ฑํ•˜๋Š” ๊ฒƒ)๋Š” ๋ชจ๋ธ์˜ ์„ฑ๋Šฅ์— ํ•ด๋ฅผ ๋ผ์น˜์ง€ ์•Š์œผ๋ฉฐ, ๋•Œ๋กœ๋Š” ์„ฑ๋Šฅ์„ ํ–ฅ์ƒ์‹œํ‚ฌ ์ˆ˜๋„ ์žˆ์Šต๋‹ˆ๋‹ค! ๊ฒฐ๊ณผ 2์˜ ์ค‘์š”์„ฑ์„ ๊ฐ•์กฐํ•ด ๋ด…์‹œ๋‹ค.
* [label=*]* ๋‹ค์–‘ํ•œ ๋ฐ์ดํ„ฐ(๋™์ผํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์—ฌ๋Ÿฌ ๋ฒˆ ๋‹ค์‹œ ์ž‘์„ฑ)๋Š” ๋ชจ๋ธ์˜ ์„ฑ๋Šฅ์— ํ•ด๋ฅผ ๋ผ์น˜์ง€ ์•Š์œผ๋ฉฐ ๋•Œ๋กœ๋Š” ๊ฐœ์„ ๋  ์ˆ˜๋„ ์žˆ์Šต๋‹ˆ๋‹ค!๊ฒฐ๊ณผ 2์˜ ์ค‘์š”์„ฑ์„ ๊ฐ•์กฐํ•ด ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.
The International Emmy Awards Competition, now in its 51st year, accepted submissions across all categories starting from December 7, 2022, until the deadline on February 16, 2023.
๊ตญ์ œ ์—๋ฏธ์ƒ ๋Œ€ํšŒ๋Š” ์˜ฌํ•ด๋กœ 51ํšŒ๋ฅผ ๋งž์ดํ•˜๋ฉฐ, 2022๋…„ 12์›” 7์ผ๋ถ€ํ„ฐ 2023๋…„ 2์›” 16์ผ๊นŒ์ง€ ๋ชจ๋“  ๋ถ€๋ฌธ์—์„œ ์ถœํ’ˆ์ž‘์„ ์ ‘์ˆ˜ํ–ˆ์Šต๋‹ˆ๋‹ค.
์˜ฌํ•ด 51ํšŒ๋ฅผ ๋งž๋Š” ๊ตญ์ œ ์—๋ฏธ์ƒ ๊ฒฝ์—ฐ๋Œ€ํšŒ๋Š” 2022๋…„ 12์›” 7์ผ๋ถ€ํ„ฐ 2023๋…„ 2์›” 16์ผ ๋งˆ๊ฐ์ผ๊นŒ์ง€ ๋ชจ๋“  ๋ถ€๋ฌธ์˜ ์ถœํ’ˆ์ž‘์„ ์ ‘์ˆ˜ํ–ˆ์Šต๋‹ˆ๋‹ค.
**All authors** provided valuable contributions to designing, analyzing, and iterating on experiments, writing and editing the paper, and generally managing the project's progress.
๋ชจ๋“  ์ €์ž๋Š” ์‹คํ—˜ ์„ค๊ณ„, ๋ถ„์„ ๋ฐ ๋ฐ˜๋ณต, ๋…ผ๋ฌธ ์ž‘์„ฑ ๋ฐ ํŽธ์ง‘, ๊ทธ๋ฆฌ๊ณ  ์ „๋ฐ˜์ ์ธ ํ”„๋กœ์ ํŠธ ์ง„ํ–‰ ๊ด€๋ฆฌ์— ๊ท€์ค‘ํ•œ ๊ธฐ์—ฌ๋ฅผ ํ–ˆ์Šต๋‹ˆ๋‹ค.
**๋ชจ๋“  ์ €์ž**๋Š” ์‹คํ—˜ ์„ค๊ณ„, ๋ถ„์„ ๋ฐ ๋ฐ˜๋ณต, ๋…ผ๋ฌธ ์ž‘์„ฑ ๋ฐ ํŽธ์ง‘, ์ผ๋ฐ˜์ ์œผ๋กœ ํ”„๋กœ์ ํŠธ ์ง„ํ–‰ ์ƒํ™ฉ ๊ด€๋ฆฌ์— ๊ท€์ค‘ํ•œ ๊ธฐ์—ฌ๋ฅผ ํ•˜์˜€์Šต๋‹ˆ๋‹ค.
1. Apply _SemDeDup_ with a selection ratio \(R_{dedup}\) on the entire dataset \(D\), producing a smaller dataset \(D^{\prime}\) 2. Cluster points in \(D^{\prime}\) with K-Means 3.
1. ์ „์ฒด ๋ฐ์ดํ„ฐ์…‹ \(D\)์— ์„ ํƒ ๋น„์œจ \(R_{dedup}\)์„ ์ ์šฉํ•˜์—ฌ _SemDeDup_์„ ์ ์šฉํ•˜๊ณ , ๋” ์ž‘์€ ๋ฐ์ดํ„ฐ์…‹ \(D^{\prime}\)์„ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค. 2. K-Means๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ \(D^{\prime}\)์˜ ํฌ์ธํŠธ๋“ค์„ ํด๋Ÿฌ์Šคํ„ฐ๋งํ•ฉ๋‹ˆ๋‹ค. 3.
1. ์ „์ฒด ๋ฐ์ดํ„ฐ ์ง‘ํ•ฉ \(D\)์— ์„ ํƒ ๋น„์œจ \(R_{dedup}\)์„ ์‚ฌ์šฉํ•˜์—ฌ _SemDeDup_์„ ์ ์šฉํ•˜์—ฌ ๋” ์ž‘์€ ๋ฐ์ดํ„ฐ ์ง‘ํ•ฉ \(D^{\prime}\)๋ฅผ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค. 2. D^{\prime}\)์˜ ํฌ์ธํŠธ๋ฅผ K-Means๋กœ ํด๋Ÿฌ์Šคํ„ฐ๋งํ•ฉ๋‹ˆ๋‹ค. 3.
Finally, while we chose various topics for the knowledge bases, all of our sources came from Wikipedia. Other datasets may yield different results, and must be evaluated carefully.
๋งˆ์ง€๋ง‰์œผ๋กœ, ์šฐ๋ฆฌ๋Š” ์ง€์‹ ๊ธฐ๋ฐ˜์„ ์œ„ํ•ด ๋‹ค์–‘ํ•œ ์ฃผ์ œ๋ฅผ ์„ ํƒํ–ˆ์ง€๋งŒ, ๋ชจ๋“  ์ถœ์ฒ˜๋Š” Wikipedia์—์„œ ๊ฐ€์ ธ์™”์Šต๋‹ˆ๋‹ค. ๋‹ค๋ฅธ ๋ฐ์ดํ„ฐ์…‹์€ ๋‹ค๋ฅธ ๊ฒฐ๊ณผ๋ฅผ ๋‚ผ ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์‹ ์ค‘ํ•˜๊ฒŒ ํ‰๊ฐ€ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
๋งˆ์ง€๋ง‰์œผ๋กœ, ์ง€์‹ ๊ธฐ๋ฐ˜์— ๋Œ€ํ•œ ๋‹ค์–‘ํ•œ ์ฃผ์ œ๋ฅผ ์„ ํƒํ–ˆ์ง€๋งŒ ๋ชจ๋“  ์†Œ์Šค๋Š” ์œ„ํ‚คํ”ผ๋””์•„์—์„œ ๊ฐ€์ ธ์™”์Šต๋‹ˆ๋‹ค. ๋‹ค๋ฅธ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ๋‹ค๋ฅธ ๊ฒฐ๊ณผ๋ฅผ ์‚ฐ์ถœํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์‹ ์ค‘ํ•˜๊ฒŒ ํ‰๊ฐ€ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
The fifty-first edition of the International Emmy Awards Competition accepted submissions across all categories from December 7, 2022, until the closing date of February 16, 2023.
์ œ51ํšŒ ๊ตญ์ œ ์—๋ฏธ์ƒ ๋Œ€ํšŒ๋Š” 2022๋…„ 12์›” 7์ผ๋ถ€ํ„ฐ 2023๋…„ 2์›” 16์ผ๊นŒ์ง€ ๋ชจ๋“  ๋ถ€๋ฌธ์—์„œ ์ถœํ’ˆ์ž‘์„ ์ ‘์ˆ˜ํ–ˆ์Šต๋‹ˆ๋‹ค.
์ œ51ํšŒ ๊ตญ์ œ ์—๋ฏธ์ƒ ๊ฒฝ์—ฐ๋Œ€ํšŒ๋Š” 2022๋…„ 12์›” 7์ผ๋ถ€ํ„ฐ 2023๋…„ 2์›” 16์ผ ๋งˆ๊ฐ์ผ๊นŒ์ง€ ๋ชจ๋“  ๋ถ€๋ฌธ์˜ ์ถœํ’ˆ์ž‘์„ ์ ‘์ˆ˜ํ–ˆ์Šต๋‹ˆ๋‹ค.
that exhibit a low FUAR value can be considered suitable for CKL. If the value is zero, then it is a case where no forgetting occurs at all and is the upper bound for performance.
FUAR ๊ฐ’์ด ๋‚ฎ์€ ๊ฒฝ์šฐ CKL์— ์ ํ•ฉํ•˜๋‹ค๊ณ  ๊ฐ„์ฃผ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ฐ’์ด 0์ธ ๊ฒฝ์šฐ, ์ „ํ˜€ ๋ง๊ฐ์ด ๋ฐœ์ƒํ•˜์ง€ ์•Š๋Š” ๊ฒฝ์šฐ๋กœ ์„ฑ๋Šฅ์˜ ์ƒํ•œ์„ ์ž…๋‹ˆ๋‹ค.
๋‚ฎ์€ FUAR ๊ฐ’์„ ๋‚˜ํƒ€๋‚ด๋Š” ๊ฒƒ์€ CKL์— ์ ํ•ฉํ•œ ๊ฒƒ์œผ๋กœ ๊ฐ„์ฃผํ•  ์ˆ˜ ์žˆ๋‹ค. ๊ฐ’์ด 0์ด๋ฉด ์žŠ์–ด๋ฒ„๋ฆผ์ด ์ „ํ˜€ ๋ฐœ์ƒํ•˜์ง€ ์•Š๋Š” ๊ฒฝ์šฐ์ด๋ฉฐ ์„ฑ๋Šฅ์˜ ์ƒํ•œ์„ ์ด๋‹ค.
- There are reports and hints from McCarthy himself that he is considering resignation and may want to "get the hell out," which could indicate a potential for an early departure.
- McCarthy ๋ณธ์ธ์œผ๋กœ๋ถ€ํ„ฐ ์‚ฌ์ž„์„ ๊ณ ๋ คํ•˜๊ณ  ์žˆ์œผ๋ฉฐ "๋นจ๋ฆฌ ๋– ๋‚˜๊ณ  ์‹ถ๋‹ค"๋Š” ์˜์‚ฌ๋ฅผ ๋‚ด๋น„์นœ ๋ณด๊ณ ์™€ ์•”์‹œ๊ฐ€ ์žˆ์–ด, ์กฐ๊ธฐ ํ‡ด์ž„ ๊ฐ€๋Šฅ์„ฑ์„ ์‹œ์‚ฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
- ๋งค์นด์‹œ ๋ณธ์ธ์˜ ๋ณด๊ณ ์™€ ํžŒํŠธ์— ๋”ฐ๋ฅด๋ฉด ์‚ฌ์ž„์„ ๊ณ ๋ คํ•˜๊ณ  ์žˆ์œผ๋ฉฐ "์ง€์˜ฅ์—์„œ ๋‚˜๊ฐ€๊ณ  ์‹ถ๋‹ค"๊ณ  ๋งํ•  ์ˆ˜ ์žˆ๋Š”๋ฐ, ์ด๋Š” ์กฐ๊ธฐ ํ‡ด์ง„ ๊ฐ€๋Šฅ์„ฑ์„ ๋‚˜ํƒ€๋‚ผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
MetaMath: According to Yu et al. (2023), GPT-3.5-Turbo is employed to rephrase queries and responses in both MATH and GSM8K datasets and to introduce reverse engineering problems.
MetaMath: Yu et al. (2023)์— ๋”ฐ๋ฅด๋ฉด, GPT-3.5-Turbo๋Š” MATH ๋ฐ GSM8K ๋ฐ์ดํ„ฐ์…‹์—์„œ ์ฟผ๋ฆฌ์™€ ์‘๋‹ต์„ ๋‹ค์‹œ ํ‘œํ˜„ํ•˜๊ณ  ์—ญ๊ณตํ•™ ๋ฌธ์ œ๋ฅผ ๋„์ž…ํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.
๋ฉ”ํƒ€๋งค์Šค: Yu ์™ธ(2023)์— ๋”ฐ๋ฅด๋ฉด, MATH ๋ฐ GSM8K ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ์ฟผ๋ฆฌ ๋ฐ ์‘๋‹ต์„ ๋ฐ”๊พธ์–ด ๋งํ•˜๊ณ  ๋ฆฌ๋ฒ„์Šค ์—”์ง€๋‹ˆ์–ด๋ง ๋ฌธ์ œ๋ฅผ ๋„์ž…ํ•˜๊ธฐ ์œ„ํ•ด GPT-3.5-Turbo๊ฐ€ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.
OptimizationWe use the identical settings as the earlier LLM PS-merging experiment in Section 4.1, and TIES-Merging with DARE for merging the source models in the parameter space.
์ตœ์ ํ™” ์šฐ๋ฆฌ๋Š” 4.1์ ˆ์˜ ์ด์ „ LLM PS-๋ณ‘ํ•ฉ ์‹คํ—˜๊ณผ ๋™์ผํ•œ ์„ค์ •์„ ์‚ฌ์šฉํ•˜๋ฉฐ, ๋งค๊ฐœ๋ณ€์ˆ˜ ๊ณต๊ฐ„์—์„œ ์†Œ์Šค ๋ชจ๋ธ์„ ๋ณ‘ํ•ฉํ•˜๊ธฐ ์œ„ํ•ด DARE์™€ ํ•จ๊ป˜ TIES-Merging์„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.
์ตœ์ ํ™”4.1์ ˆ์˜ ์ด์ „ LLM PS-๋ณ‘ํ•ฉ ์‹คํ—˜๊ณผ ๋™์ผํ•œ ์„ค์ •์„ ์‚ฌ์šฉํ•˜๊ณ , ํŒŒ๋ผ๋ฏธํ„ฐ ๊ณต๊ฐ„์—์„œ ์†Œ์Šค ๋ชจ๋ธ์„ ๋ณ‘ํ•ฉํ•˜๊ธฐ ์œ„ํ•ด DARE์™€ ํ•จ๊ป˜ TIES-Merging์„ ์‚ฌ์šฉํ•œ๋‹ค.
This demonstrates the unique value of our merged model, which effectively combines mathematical reasoning capabilities with Japanese language understanding and cultural knowledge.
์ด๋Š” ์ˆ˜ํ•™์  ์ถ”๋ก  ๋Šฅ๋ ฅ๊ณผ ์ผ๋ณธ์–ด ์ดํ•ด ๋ฐ ๋ฌธํ™” ์ง€์‹์„ ํšจ๊ณผ์ ์œผ๋กœ ๊ฒฐํ•ฉํ•œ ์šฐ๋ฆฌ์˜ ํ†ตํ•ฉ ๋ชจ๋ธ์˜ ๋…ํŠนํ•œ ๊ฐ€์น˜๋ฅผ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
์ด๋Š” ์ˆ˜ํ•™์  ์ถ”๋ก  ๋Šฅ๋ ฅ๊ณผ ์ผ๋ณธ์–ด ์ดํ•ด ๋ฐ ๋ฌธํ™”์  ์ง€์‹์„ ํšจ๊ณผ์ ์œผ๋กœ ๊ฒฐํ•ฉํ•œ ์šฐ๋ฆฌ์˜ ๋ณ‘ํ•ฉ ๋ชจ๋ธ๋งŒ์˜ ๊ณ ์œ ํ•œ ๊ฐ€์น˜๋ฅผ ๋ณด์—ฌ์ค€๋‹ค.