Image de couverture de livre

The Detection-Extraction Gap: Models Know the Answer Before They Can Say It

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Wang, Hanyang, Zhu, Mingxuan
Format:	Preprint
Publié:	2026
Sujets:	Computation and Language Artificial Intelligence Information Theory Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Know Your Limits: Entropy Estimation Modeling for Compression and Generalization
par: Badger, Benjamin L., et autres
Publié: (2025)

NanoKnow: How to Know What Your Language Model Knows
par: Gu, Lingwei, et autres
Publié: (2026)

How Many Features Can a Language Model Store Under the Linear Representation Hypothesis?
par: Garg, Nikhil, et autres
Publié: (2026)

A Survey on Large Language Models from Concept to Implementation
par: Wang, Chen, et autres
Publié: (2024)

When Models Know More Than They Say: Probing Analogical Reasoning in LLMs
par: McGovern, Hope, et autres
Publié: (2026)

Diff-eRank: A Novel Rank-Based Metric for Evaluating Large Language Models
par: Wei, Lai, et autres
Publié: (2024)

Language Modeling Is Compression
par: Delétang, Grégoire, et autres
Publié: (2023)

The Information of Large Language Model Geometry
par: Tan, Zhiquan, et autres
Publié: (2024)

Geometric Signatures of Compositionality Across a Language Model's Lifetime
par: Lee, Jin Hwa, et autres
Publié: (2024)

SQuat: Subspace-orthogonal KV Cache Quantization
par: Wang, Hao, et autres
Publié: (2025)

Subjective Depth and Timescale Transformers: Learning Where and When to Compute
par: Wieser, Frederico, et autres
Publié: (2025)

From Data to Behavior: Predicting Unintended Model Behaviors Before Training
par: Wang, Mengru, et autres
Publié: (2026)

Explore BiLSTM-CRF-Based Models for Open Relation Extraction
par: Ni, Tao, et autres
Publié: (2021)

Diffusion Language Models Know the Answer Before Decoding
par: Li, Pengxiang, et autres
Publié: (2025)

Multilingual Non-Factoid Question Answering with Answer Paragraph Selection
par: Mishra, Ritwik, et autres
Publié: (2024)

Small Encoders Can Rival Large Decoders in Detecting Groundedness
par: Abbes, Istabrak, et autres
Publié: (2025)

Towards A Unified View of Answer Calibration for Multi-Step Reasoning
par: Deng, Shumin, et autres
Publié: (2023)

Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain
par: Liu, Wei, et autres
Publié: (2026)

Analyzing and Improving Chain-of-Thought Monitorability Through Information Theory
par: Anwar, Usman, et autres
Publié: (2026)

A Communication-Theoretic Framework for LLM Agents: Cost-Aware Adaptive Reliability
par: Omidvar, Hamed, et autres
Publié: (2026)

The Stepwise Informativeness Assumption: Why are Entropy Dynamics and Reasoning Correlated in LLMs?
par: Català, Mar Gonzàlez I, et autres
Publié: (2026)

Learning is Forgetting: LLM Training As Lossy Compression
par: Conklin, Henry C., et autres
Publié: (2026)

A Training-free Method for LLM Text Attribution
par: Radvand, Tara, et autres
Publié: (2025)

Optimal Quantization for Matrix Multiplication
par: Ordentlich, Or, et autres
Publié: (2024)

Memorization-Compression Cycles Improve Generalization
par: Yu, Fangyuan
Publié: (2025)

Compression Represents Intelligence Linearly
par: Huang, Yuzhen, et autres
Publié: (2024)

Measuring Uncertainty in Transformer Circuits with Effective Information Consistency
par: Krasnovsky, Anatoly A.
Publié: (2025)

SPEX: Scaling Feature Interaction Explanations for LLMs
par: Kang, Justin Singh, et autres
Publié: (2025)

An Information Theoretic Perspective on Agentic System Design
par: He, Shizhe, et autres
Publié: (2025)

Think Before you Write: QA-Guided Reasoning for Character Descriptions in Books
par: Papoudakis, Argyrios, et autres
Publié: (2026)

Few-shot Continual Relation Extraction via Open Information Extraction
par: Nguyen, Thiem, et autres
Publié: (2025)

FaKnow: A Unified Library for Fake News Detection
par: Zhu, Yiyuan, et autres
Publié: (2024)

Verif.ai: Towards an Open-Source Scientific Generative Question-Answering System with Referenced and Verifiable Answers
par: Košprdić, Miloš, et autres
Publié: (2024)

KnowCoder-X: Boosting Multilingual Information Extraction via Code
par: Zuo, Yuxin, et autres
Publié: (2024)

GLiNER multi-task: Generalist Lightweight Model for Various Information Extraction Tasks
par: Stepanov, Ihor, et autres
Publié: (2024)

Reflect then Learn: Active Prompting for Information Extraction Guided by Introspective Confusion
par: Zhao, Dong, et autres
Publié: (2025)

Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking
par: Zelikman, Eric, et autres
Publié: (2024)

SelfReflect: Can LLMs Communicate Their Internal Answer Distribution?
par: Kirchhof, Michael, et autres
Publié: (2025)

LongKey: Keyphrase Extraction for Long Documents
par: Alves, Jeovane Honorio, et autres
Publié: (2024)

The Shape of Learning: Anisotropy and Intrinsic Dimensions in Transformer-Based Models
par: Razzhigaev, Anton, et autres
Publié: (2023)