Copertina

Epistemic Traps: Rational Misalignment Driven by Model Misspecification

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Xu, Xingcheng, Qu, Jingjing, Zhang, Qiaosheng, Lu, Chaochao, Yang, Yanqing, Zou, Na, Hu, Xia
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Artificial Intelligence Computation and Language Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

KALE: Enhancing Knowledge Manipulation in Large Language Models via Knowledge-aware Learning
di: Lv, Qitan, et al.
Pubblicazione: (2026)

RAPO: Risk-Aware Preference Optimization for Generalizable Safe Reasoning
di: Wei, Zeming, et al.
Pubblicazione: (2026)

SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces
di: Jin, Chang, et al.
Pubblicazione: (2026)

MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety
di: Wen, Xiaoyu, et al.
Pubblicazione: (2026)

The Policy Cliff: A Theoretical Analysis of Reward-Policy Maps in Large Language Models
di: Xu, Xingcheng
Pubblicazione: (2025)

Dynamic Adversarial Reinforcement Learning for Robust Multimodal Large Language Models
di: Bao, Yicheng, et al.
Pubblicazione: (2026)

Principled Understanding of Generalization for Generative Transformer Models in Arithmetic Reasoning Tasks
di: Xu, Xingcheng, et al.
Pubblicazione: (2024)

It Ain't That Bad: Understanding the Mysterious Performance Drop in OOD Generalization for Generative Transformer Models
di: Xu, Xingcheng, et al.
Pubblicazione: (2023)

Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking
di: He, Zhida, et al.
Pubblicazione: (2026)

TrinityGuard: A Unified Framework for Safeguarding Multi-Agent Systems
di: Wang, Kai, et al.
Pubblicazione: (2026)

CauScientist: Teaching LLMs to Respect Data for Causal Discovery
di: Peng, Bo, et al.
Pubblicazione: (2026)

Emergent and Subliminal Misalignment Through the Lens of Data-Mediated Transfer
di: Askin, Baris, et al.
Pubblicazione: (2026)

Can Post-Training Transform LLMs into Causal Reasoners?
di: Chen, Junqi, et al.
Pubblicazione: (2026)

Are More Tokens Rational? Inference-Time Scaling in Language Models as Adaptive Resource Rationality
di: Hu, Zhimin, et al.
Pubblicazione: (2026)

RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation
di: Liang, Kaiqu, et al.
Pubblicazione: (2025)

Thought Crime: Backdoors and Emergent Misalignment in Reasoning Models
di: Chua, James, et al.
Pubblicazione: (2025)

RePO: Replay-Enhanced Policy Optimization
di: Li, Siheng, et al.
Pubblicazione: (2025)

Ulterior Motives: Detecting Misaligned Reasoning in Continuous Thought Models
di: Ramjee, Sharan
Pubblicazione: (2026)

Towards AI-$45^{\circ}$ Law: A Roadmap to Trustworthy AGI
di: Yang, Chao, et al.
Pubblicazione: (2024)

Persona-Model Collapse in Emergent Misalignment
di: Costa, Davi Bastos, et al.
Pubblicazione: (2026)

CLEAR: Can Language Models Really Understand Causal Graphs?
di: Chen, Sirui, et al.
Pubblicazione: (2024)

ESI: Epistemic Uncertainty Quantification via Semantic-preserving Intervention for Large Language Models
di: Li, Mingda, et al.
Pubblicazione: (2025)

Decoupled Reasoning with Implicit Fact Tokens (DRIFT): A Dual-Model Framework for Efficient Long-Context Inference
di: Xie, Wenxuan, et al.
Pubblicazione: (2026)

Beyond Surface Structure: A Causal Assessment of LLMs' Comprehension Ability
di: Han, Yujin, et al.
Pubblicazione: (2024)

Causal Evaluation of Language Models
di: Chen, Sirui, et al.
Pubblicazione: (2024)

Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models
di: Hu, Zhibo, et al.
Pubblicazione: (2026)

Can We Predict Alignment Before Models Finish Thinking? Towards Monitoring Misaligned Reasoning Models
di: Chan, Yik Siu, et al.
Pubblicazione: (2025)

MAVEN: Multi-Agent Verification-Elaboration Network with In-Step Epistemic Auditing
di: Yao, Yinsheng, et al.
Pubblicazione: (2026)

A Neuro-inspired Interpretation of Unlearning in Large Language Models through Sample-level Unlearning Difficulty
di: Feng, Xiaohua, et al.
Pubblicazione: (2025)

Rational Metareasoning for Large Language Models
di: De Sabbata, C. Nicolò, et al.
Pubblicazione: (2024)

Interactive Training: Feedback-Driven Neural Network Optimization
di: Zhang, Wentao, et al.
Pubblicazione: (2025)

Unintended Misalignment from Agentic Fine-Tuning: Risks and Mitigation
di: Hahm, Dongyoon, et al.
Pubblicazione: (2025)

Does Visual Rendering Bypass Tokenization? Investigating Script-Tokenizer Misalignment in Pixel-Based Language Models
di: Susanto, Lucky, et al.
Pubblicazione: (2026)

The Devil in the Details: Emergent Misalignment, Format and Coherence in Open-Weights LLMs
di: Dickson, Craig
Pubblicazione: (2025)

Solving Inequality Proofs with Large Language Models
di: Lu, Pan, et al.
Pubblicazione: (2025)

The Flexibility Trap: Why Arbitrary Order Limits Reasoning Potential in Diffusion Language Models
di: Ni, Zanlin, et al.
Pubblicazione: (2026)

Bridging Draft Policy Misalignment: Group Tree Optimization for Speculative Decoding
di: Hu, Shijing, et al.
Pubblicazione: (2025)

Integration of Large Language Models and Federated Learning
di: Chen, Chaochao, et al.
Pubblicazione: (2023)

Epistemic Observability in Language Models
di: Mason, Tony, et al.
Pubblicazione: (2026)

More Bang for the Buck: Process Reward Modeling with Entropy-Driven Uncertainty
di: Cao, Lang, et al.
Pubblicazione: (2025)