AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

DARE-bench: Evaluating Modeling and Instruction Fidelity of LLMs in Data Science

Fan Shu, Yite Wang, Ruofan Wu, Boyi Liu, Zhewei Yao, Yuxiong He, Feng Yan

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning systems · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add MM data to IF data in SFT-DV training (ablation)Qwen3-32B SFT-DV · DARE-bench class-MMscore: 22.38 -> 43.82Improved
add IF data to MM data in SFT-DV training (ablation)Qwen3-32B SFT-DV · DARE-bench class-IFscore: 11.71 -> 38.58Improved
train on MM data only instead of baseline (ablation)Qwen3-32B SFT-DV · DARE-bench class-IFscore: 17.11 -> 11.71Hurt
train on IF data only instead of baseline (ablation)Qwen3-32B SFT-DV · DARE-bench class-MMscore: 30.71 -> 22.38Hurt
SFT with all-valid rejection samplingQwen3-32B · DARE-bench Totaltotal score: 23.25 -> 42.91Improved
SFT with Duo-Valid rejection samplingQwen3-32B · DSBenchaccuracy: 32.38 -> 42.41Improved
use inclusive sampling (DV) instead of strict filtering (FV) (ablation)Qwen3-32B SFT · DSBenchaccuracy: 37.82 -> 42.41Improved
use Fastest-Valid instead of All-Valid sampling (ablation)Qwen3-32B SFT · DARE-bench Totaltotal score: 42.91 -> 42.42Mixed
GRPO reinforcement learningQwen3-4B · DARE-bench Totaltotal score: 4.39 -> 37.40Improved
GRPO reinforcement learningQwen3-4B · DSBenchaccuracy: 18.23 -> 40.00Improved
increase interactive turns from 3 to 5 (time 300 s to 200 s) (ablation)o4-mini · DARE-bench class-IFscore: 37.16 -> 67.56Improved
increase interactive turns from 5 to 15 (time 200 s to 100 s) (ablation)o4-mini · DARE-bench class-IFscore: 67.56 -> 76.80Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy