Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection
Paper recorded by Signals 4 on 2026-10-09 in cs.CL. Abstract reproduced from arXiv; link to the original below.
Published 2026-10-09 on arXiv · recorded by Signals 4 on 2026-10-12
Category: cs.CL · 自然语言处理 · first seen 2026-10-12
Abstract
A useful language model should know when to answer and when to say "I don't know." Can we improve that decision by training the model itself or by training a separate component to check its answers? We study both approaches on Qwen3-14B for factual questions, using a score that rewards correct answers, penalizes mistakes, and allows abstention. Across three training seeds and two held-out confiden
Read on arXiv →
Cite this page: Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection: the #269 most recent of 660 cs.CL papers we have recorded (as of 2026-10-09). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/can-language-models-learn-when-to-abstain-policy-learning-and-calibrated-answer-.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable:
papers.json
Get 4 AI signals a day by email — free.
Get 4 AI signals a day by email — free