AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

LFQA-E: Carefully Benchmarking Long-form QA Evaluation

Yuchen Fan, Chen Ling, Xin Zhong, Shuo Zhang, Heng Zhou, Yuchen Zhang, Mingyu Liang, Chengxing Xie, Ermo Hua, Zhizhou He, Cheng Huang, Gang Chen and 2 more

ICLR 2026 · Large language models › Retrieval and knowledge · approach: Explanatory and multiple-choice QA benchmarks · 1 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
scale from Qwen2.5-32B to Qwen2.5-72B as evaluatorQwen2.5 evaluator · LFQA-E avgaccuracy (%): 60.1 -> 57.1Hurt
use a reasoning model (o1-mini) instead of GPT-4oLLM evaluator · LFQA-E avgaccuracy (%): 57.5 -> 60.9Improved
set evaluator temperature to 0 for o1-mini (ablation)o1-mini · LFQA-E-zhaccuracy (%): 58.9 -> 5.8Hurt
set evaluator temperature to 0 for GPT-4o (ablation)GPT-4o · LFQA-E avgmacro-F1 (%): 44.5 -> 46.7Improved
remove CoT, ask for final answer only (ablation)Qwen2.5-32B-Instruct · LFQA-E-enaccuracy (%): 63.5 -> 57.3Hurt
remove CoT, ask for final answer only (ablation)DeepSeek-R1 · LFQA-E-zhaccuracy (%): 57.8 -> 60.0Mixed
drop tie-labeled comparisons (ablation)GPT-4o · LFQA-E avgaccuracy (%): 57.5 -> 63.9Improved
add TTRL test-time RL trainingQwen2.5-7B-Instruct · LFQA-E-enaccuracy (%): 60.6 -> 68.2Improved
add Clip Higher to TTRL (ablation)Qwen2.5-3B-Instruct · LFQA-E-enaccuracy (%): 63.9 -> 66.5Improved
add Clip Higher to TTRL (ablation)Qwen2.5-7B-Instruct · LFQA-E-enaccuracy (%): 68.2 -> 68.6No clear effect
use structured answer-tag prompt instead of CoT (ablation)Qwen2.5-3B-Instruct · LFQA-E-enaccuracy (%): 49.6 -> 59.7Improved
use length as metric instead of reference-based judgingLength · LFQA-E avgaccuracy (%): 60.1 -> 47.7Hurt

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