Image de couverture de livre

Biasless Language Models Learn Unnaturally: How LLMs Fail to Distinguish the Possible from the Impossible

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Ziv, Imry, Lan, Nur, Chemla, Emmanuel
Format:	Preprint
Publié:	2025
Sujets:	Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Large Language Models as Proxies for Theories of Human Linguistic Cognition
par: Ziv, Imry, et autres
Publié: (2025)

Bridging the Empirical-Theoretical Gap in Neural Network Formal Language Learning Using Minimum Description Length
par: Lan, Nur, et autres
Publié: (2024)

What Makes Two Language Models Think Alike?
par: Salle, Jeanne, et autres
Publié: (2024)

A Minimum Description Length Approach to Regularization in Neural Networks
par: Abudy, Matan, et autres
Publié: (2025)

Unnatural Languages Are Not Bugs but Features for LLMs
par: Duan, Keyu, et autres
Publié: (2025)

No Such Thing as a General Learner: Language models and their dual optimization
par: Chemla, Emmanuel, et autres
Publié: (2024)

Shades of Zero: Distinguishing Impossibility from Inconceivability
par: Hu, Jennifer, et autres
Publié: (2025)

Improving Spoken Language Modeling with Phoneme Classification: A Simple Fine-tuning Approach
par: Poli, Maxime, et autres
Publié: (2024)

Metric Learning Encoding Models: A Multivariate Framework for Interpreting Neural Representations
par: Jalouzot, Louis, et autres
Publié: (2024)

Probing Syntax in Large Language Models: Successes and Remaining Challenges
par: Diego-Simón, Pablo J., et autres
Publié: (2025)

Polar probe linearly decodes semantic structures from LLMs
par: Diego-Simón, Pablo J., et autres
Publié: (2026)

fastabx: A library for efficient computation of ABX discriminability
par: Poli, Maxime, et autres
Publié: (2025)

The Impact of Syntactic and Semantic Proximity on Machine Translation with Back-Translation
par: Guerin, Nicolas, et autres
Publié: (2024)

A Neural Model for Word Repetition
par: Dager, Daniel, et autres
Publié: (2025)

How LLMs Fail to Support Fact-Checking
par: Proma, Adiba Mahbub, et autres
Publié: (2025)

The Impossibility of Fair LLMs
par: Anthis, Jacy, et autres
Publié: (2024)

Rolling the DICE on Idiomaticity: How LLMs Fail to Grasp Context
par: Mi, Maggie, et autres
Publié: (2024)

Mission: Impossible Language Models
par: Kallini, Julie, et autres
Publié: (2024)

Learning from Mistakes: Self-correct Adversarial Training for Chinese Unnatural Text Correction
par: Feng, Xuan, et autres
Publié: (2024)

Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
par: Roy, Amartya, et autres
Publié: (2026)

When Attention Collapses: How Degenerate Layers in LLMs Enable Smaller, Stronger Models
par: Sanyal, Sunny, et autres
Publié: (2024)

Why Gaussian Diffusion Models Fail on Discrete Data and How to Prevent It?
par: Shabalin, Alexander, et autres
Publié: (2026)

Why We Need World Models for AGI: Where LLMs Fail and How World Models May Outperform
par: Alaswad, Feisal, et autres
Publié: (2026)

ImpossibleBench: Measuring LLMs' Propensity of Exploiting Test Cases
par: Zhong, Ziqian, et autres
Publié: (2025)

Compositional Bias Control in Large Language Models: Preference Learning Fails, Supervision Succeeds
par: Mahesh, Atij
Publié: (2025)

The Validation Gap: A Mechanistic Analysis of How Language Models Compute Arithmetic but Fail to Validate It
par: Bertolazzi, Leonardo, et autres
Publié: (2025)

Mission Impossible: A Statistical Perspective on Jailbreaking LLMs
par: Su, Jingtong, et autres
Publié: (2024)

Tracking the Limits of Knowledge Propagation: How LLMs Fail at Multi-Step Reasoning with Conflicting Knowledge
par: Feng, Yiyang, et autres
Publié: (2026)

Language Models Fail to Introspect About Their Knowledge of Language
par: Song, Siyuan, et autres
Publié: (2025)

Lost in the Noise: How Reasoning Models Fail with Contextual Distractors
par: Lee, Seongyun, et autres
Publié: (2026)

On the Fundamental Impossibility of Hallucination Control in Large Language Models
par: Karpowicz, Michał P.
Publié: (2025)

A polar coordinate system represents syntax in large language models
par: Diego-Simón, Pablo, et autres
Publié: (2024)

Large Language Models and Impossible Language Acquisition: "False Promise" or an Overturn of our Current Perspective towards AI
par: Wang, Ziyan, et autres
Publié: (2026)

Distinguishing the Knowable from the Unknowable with Language Models
par: Ahdritz, Gustaf, et autres
Publié: (2024)

When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs
par: Li, Xiaomin, et autres
Publié: (2025)

Retrieving Evidence from EHRs with LLMs: Possibilities and Challenges
par: Ahsan, Hiba, et autres
Publié: (2023)

Why Supervised Fine-Tuning Fails to Learn: A Systematic Study of Incomplete Learning in Large Language Models
par: Xue, Chao, et autres
Publié: (2026)

From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning
par: Shani, Chen, et autres
Publié: (2025)

Mission Impossible: Feedback-Guided Dynamic Interactive Planning for Improving Reasoning on LLMs
par: Yan, Dong, et autres
Publié: (2025)

DiscoPhon: Benchmarking the Unsupervised Discovery of Phoneme Inventories With Discrete Speech Units
par: Poli, Maxime, et autres
Publié: (2026)