Selective QA

QA: Question Answering

Follow this topic Ask about this topic

Latest papers

10 papers2 papers in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.

Training Negatives Still Matter After Calibrating Answerability Readouts
Oct 19, 2026Jinchen Zhang, Zheyuan Liu, Yihan Zhu +1Selective QA Language Model Calibration
University of Notre Dame
Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection
Oct 9, 2026Juan Terven, Diana-Margarita Córdova-Esparza, Ruben Gomez +3Selective QA LLM Abstention
CICATA Qro., Instituto Politécnico Nacional, México · Facultad de Informática, Universidad Autónoma de Querétaro, México
Before Answering: Evidence Sufficiency under Size-Matched Memory Construction
Sep 26, 2026Joyanta Jyoti Mondal, Md. Shifatul Ahsan Apurba, Mridul Banik +2Selective QA Memory-Augmented Agents
Department of Computer and Information Sciences, University of Delaware, USA · Department of Biomedical Informatics and Data Science, University of Alabama at Birmingham, USA · Luddy School of Informatics, Computing, and Engineering, Indiana University Indianapolis, USA +2
Cite or Decline: A Strict Course-Grounded Chatbot for STEM Lecture Videos
Sep 1, 2026S M Masrur Ahmed, Jaspal SubhlokEvidence Retrieval Evidence-Grounded Generation
University of Houston
Learning Evidence Sufficiency Boundaries for Selective Answering in Grounded Multi-Hop QA
Sep 1, 2026Haruto Sato, Yuki Tanaka, Ren Nakamura +2Multi-Hop QA Evidence-Grounded Reasoning
Independent Researcher
Asymptotic Risk Calibration for Selective Question Answering
Aug 12, 2026Shufan Lin, Sijin DongQuestion Answering Selective QA
Zhangjiang University · Ibaraki University
To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning
Jul 12, 2026Fengji Zhang, Tianyu Fan, Yuxiang Zheng +4Question Answering RL for LLM Agents
City University of Hong Kong · Alibaba Group
Selective QA over Conflicting Multi-Source Personal Memory: A Diagnostic Testbed and Method Comparison
May 28, 2026Tiancheng Yang, Matthias Schonlau, Ilia SucholutskyConversational Memory Question Answering
SURE-RAG: Sufficiency and Uncertainty-Aware Evidence Verification for Selective Retrieval-Augmented Generation
May 5, 2026Jingxi Qiu, Zeyu Han, Cheng HuangRetrieval-Augmented Generation LLM Answer Verification
ZenWeave AI · Georgetown University · Southern Methodist University
PassiveQA: A Three-Action Framework for Epistemically Calibrated Question Answering via Supervised Finetuning
Apr 6, 2026Madhav S BaidyaSelective QA Question Answering
Indian Institute of Technology (BHU) Varanasi

Oct 19, 2026J/K move · Enter open · S save

Training Negatives Still Matter After Calibrating Answerability Readouts

Jinchen Zhang, Zheyuan Liu, Yihan Zhu, Meng Jiang

University of Notre Dame

Selective QA Language Model Calibration

Open paper AskPDF arXiv

Answerability readouts predict whether a passage supplies an answer, but whether calibration can compensate for removing negative passages during training remains unclear. We present a controlled study of how training-negative construction affects calibrated prediction, ranking, and downstream answer retention. We compare native training, which retains negative passages, with empty training, which removes them, holding questions, labels, positive examples, and model weights fixed. Both are calibrated on the same labeled native examples. On SQuAD 2.0 and RepLiQA, native-trained logistic readouts at selected depths achieve lower Brier loss for Qwen2.5-7B-Instruct and Gemma-2-9B-it, with or without output scores. Native fusion also outperforms calibrated output alone. Post-hoc AUROC analysis shows better ranking under native training, beyond what threshold adjustment alone can explain. On the official SQuAD 2.0 development set, native readout policies improve F1 over empty policies but retain more answers to unanswerable questions. The advantage varies by estimator and depth, while RepLiQA's text shortcuts limit claims about contextual understanding. Unfinished generations prevent complete answer-level evaluation, although worst-case bounds preserve the native-empty F1 gains. Our findings show that training negatives affect ranking and downstream decisions even after common calibration, without establishing a universal advantage for native training.