Francesco Croce, Maksym Andriushchenko, Nicolas Flammarion, Hao Zhao
We lifted 2 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| tml-epfl/long-is-more-for-alignment | canonical | 1 of 1 |
| copy not recorded | — | 1 of 1 |
| Function | Status | Where it lives |
|---|---|---|
| gen_prompt | Ran | tml-epfl/long-is-more-for-alignment/evaluation/evaluation_gpt4.py pointer only (licence: NONE) · get_code("36c101b773eb28ed") |
| parse_score | Ran | this paper's copy was not recorded; identical code first harvested from artidoro/qlora pointer only · get_code("8049b382893c73dd") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
There is a consensus that instruction fine-tuning of LLMs requires high-quality data, but what are they? LIMA (NeurIPS 2023) and AlpaGasus (ICLR 2024) are state-of-the-art methods for selecting such high-quality examples, either via manual curation or using GPT-3.5-Turbo as a quality scorer. We show that the extremely simple baseline of selecting the 1,000 instructions with longest responses-that intuitively contain more learnable information and are harder to overfitfrom standard datasets can consistently outperform these sophisticated methods according to GPT-4 and PaLM-2 as judges, while remaining competitive on the Open LLM benchmarks that test factual knowledge. We demonstrate this for several LLMs (Llama-2-7B, Llama-2-13B, Mistral-7B-v0.1) and datasets (Alpaca-52k, Evol-Instruct-70k). In addition, a lightweight refinement of such long instructions can further improve the abilities of the fine-tuned LLMs, and allows us to obtain competitive results on MT-Bench and the 2nd highest-ranked Llama-2-7B-based model on AlpacaEval 2.0, while training on only 1,000 examples and no extra preference data. We also conduct a thorough analysis of our models to ensure that their enhanced performance is not simply due to GPT-4's preference for longer responses. Overall, our findings suggest that fine-tuning on the longest responses should be the default baseline for any work on instruction fine-tuning. We provide our code in this GitHub repository.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2402.04833")
get_code_for_paper("2402.04833")
have("2402.04833")
Connect an agent — have() is free.