·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Generalizing Beyond Suboptimality: Offline Reinforcement Learning Learns Effective Scheduling through Random Solutions47m◆MARIC: Multi-Agent Reasoning for Image Classification47m◆The Impossibility of Eliciting Latent Knowledge47m◆A Five-Plane Reference Architecture for Runtime Governance of Production AI Agents47m◆PROJECTMEM: A Local-First, Event-Sourced Memory and Judgment Layer for AI Coding Agents47m◆Nonslop: A Gamified Experiment in Human-AI Collaborative Writing47m◆Geometric Metrics and LLMs: What They Measure and When They Work47m◆From Architecture to Output: Structural Origins of Hallucination in Large Language Models and the Amplifying Role of Data47m◆From Consumption to Reflection: Designing Human-AI Relations for Stable Reasoning47m◆PoQ-Judge: A Multi-Architecture Evaluation Framework for Cost-Aware Proof-of-Quality in Decentralized LLM Inference47m◆MA-DLE: Speech-based Automatic Depression Level Estimation via Memory Augmentation47m◆The Structural Attention Tax: How Retrieval Format Hijacks In-Context Learning Independent of Content47m◆Noise-Guided Transport for Imitation Learning47m◆NightFeats @ MMU-RAGent NeurIPS 2025: A Context-Optimized Multi-Agent RAG System for the Text-to-Text Track47m◆To Intervene or Not: Guiding Inference-time Alignment with Probabilistic Model Blending47m◆Dual-Stance Evaluation of Sycophancy: The Structure of Agreement and the Limits of Intervention47m◆BioDivergence: A Benchmark and Evaluation Framework for Hidden Contextual Contradictions in Biomedical Abstracts47m◆When Researchers Say Mental Model/Theory of Mind of AI, What Are They Really Talking About?47m◆ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward47m◆T2MM: An LLM Supported Architecture For Inquiry-Based Modeling47m◆Generalizing Beyond Suboptimality: Offline Reinforcement Learning Learns Effective Scheduling through Random Solutions47m◆MARIC: Multi-Agent Reasoning for Image Classification47m◆The Impossibility of Eliciting Latent Knowledge47m◆A Five-Plane Reference Architecture for Runtime Governance of Production AI Agents47m◆PROJECTMEM: A Local-First, Event-Sourced Memory and Judgment Layer for AI Coding Agents47m◆Nonslop: A Gamified Experiment in Human-AI Collaborative Writing47m◆Geometric Metrics and LLMs: What They Measure and When They Work47m◆From Architecture to Output: Structural Origins of Hallucination in Large Language Models and the Amplifying Role of Data47m◆From Consumption to Reflection: Designing Human-AI Relations for Stable Reasoning47m◆PoQ-Judge: A Multi-Architecture Evaluation Framework for Cost-Aware Proof-of-Quality in Decentralized LLM Inference47m◆MA-DLE: Speech-based Automatic Depression Level Estimation via Memory Augmentation47m◆The Structural Attention Tax: How Retrieval Format Hijacks In-Context Learning Independent of Content47m◆Noise-Guided Transport for Imitation Learning47m◆NightFeats @ MMU-RAGent NeurIPS 2025: A Context-Optimized Multi-Agent RAG System for the Text-to-Text Track47m◆To Intervene or Not: Guiding Inference-time Alignment with Probabilistic Model Blending47m◆Dual-Stance Evaluation of Sycophancy: The Structure of Agreement and the Limits of Intervention47m◆BioDivergence: A Benchmark and Evaluation Framework for Hidden Contextual Contradictions in Biomedical Abstracts47m◆When Researchers Say Mental Model/Theory of Mind of AI, What Are They Really Talking About?47m◆ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward47m◆T2MM: An LLM Supported Architecture For Inquiry-Based Modeling47m◆
News/Residual Drift Dominates Contradiction in Multi-Turn Constraint Reasoning
arxiv
PublishedMay 26, 2026 at 4:00 AM
—neutral

Residual Drift Dominates Contradiction in Multi-Turn Constraint Reasoning

Source
arxiv.orgfull article ↗
Read on arxiv→
Publisher summary· verbatim

arXiv:2605.23940v1 Announce Type: new Abstract: How do multi-turn reasoning systems fail? The expected answer is logical contradiction, in which the system's maintained state becomes unsatisfiable. We show that the dominant mode is instead satisfiable drift, where the internal state stays consistent

Stay posted· Newsletter

A 5-min weekly brief — top movers, price watch, story of the week.

// no spam · unsubscribe one-click · free forever

Discussion
Mentioned models
01
  • 01
    MUS-Repair
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#reasoning#benchmark#multi-turn#safety

No replies yet. Be first.

Mentioned models
01
  • 01
    MUS-Repair
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#reasoning#benchmark#multi-turn#safety

Related coverage

More from ARXIV
arxivGeneralizing Beyond Suboptimality: Offline Reinforcement Learning Learns Effective Scheduling through Random Solutions47marxivMARIC: Multi-Agent Reasoning for Image Classification47marxivThe Impossibility of Eliciting Latent Knowledge47marxivA Five-Plane Reference Architecture for Runtime Governance of Production AI Agents47m
The Bubble Brief
WEEKLY

Read reasoning insights every Tuesday — top movers, new releases, story of the week.

// no spam · unsubscribe one-click · free forever

Originally published on arxiv ↗
HomeModelsNews