AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

When Silence Is Golden: Can LLMs Learn to Abstain in Temporal QA and Beyond?

Xinyu Zhou, Chang Jin, Carsten Eickhoff, Zhijiang Guo, Seyed Ali Bahrainian

ICLR 2026 · Large language models › Retrieval and knowledge · approach: Explanatory and multiple-choice QA benchmarks · 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 CoT-SFT cold start then GRPO RL with abstention-aware rewardQwen2.5-1.5B-Instruct · TimeQA-HardExact Match (%): 29.95 -> 35.75 (vs GPT-4o)Improved
use time-relevant sub-context instead of original context (ablation)GPT-4o · TimeQA-HardExact Match (%): 29.95 -> 34.12Improved
use time-relevant sub-context instead of original context (ablation)GPT-3.5 · TimeQA-HardExact Match (%): 21.09 -> 18.52Hurt
provide knowledge graph facts instead of the original context (ablation)GPT-4o · TimeQA-EasyExact Match (%): 39.95 -> 30.10Hurt
extract KGs from whole context instead of chunked context (ablation)GPT-4o · TimeQA-EasyExact Match (%): 24.66 -> 30.10Improved
supervised fine-tuning on contextQwen2.5-1.5B-Instruct · TimeQA-EasyExact Match (%): 8.06 -> 1.81Hurt
add original context to question-only prompt (ablation)Qwen2.5-7B-Instruct · TimeQA-HardExact Match (%): 11.69 -> 9.00Hurt
use ROUGE+EM answer reward instead of ROUGE only (ablation)RL, Qwen2.5-1.5B-Instruct · TimeQA-HardExact Match (%): 26.67 -> 35.75Improved
add BERTScore to the ROUGE+EM reward (ablation)RL, Qwen2.5-1.5B-Instruct · TimeQA-Easy vs HardExact Match (%): Easy 43.41 -> 33.71; Hard 35.75 -> 37.62Mixed
raise KL coefficient from 0.01 to 0.04 (ablation)RL, Qwen2.5-1.5B-Instruct · TimeQA-EasyExact Match (%): 42.74 -> 41.56Hurt
increase unanswerable share in RL training data to 50% (ablation)RL, Qwen2.5-1.5B-Instruct · TimeQA-EasyExact Match (%): 43.41 -> 11.24Hurt
use LoRA instead of full-parameter SFT (ablation)Qwen2.5-1.5B-Instruct · TimeQA-EasyFalse positives (count): 309 -> 437Hurt

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