·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
MuQ-Eval: An Open-Source Per-Sample Quality Metric for AI Music Generation Evaluation8h◆ThinkJev: One Thought, Every Budget for Calibrated Typed Decisions8h◆What Does an Ontology Catch? Curated, LLM-Generated and Data-Derived Ontologies as Checkers of LLM Agents over Knowledge Graphs8h◆GAUSS: A Workload-Aware Analytical and Monte-Carlo Simulator for LLM Inference Serving8h◆Algorithmic Bias from the Perspectives of Healthcare Professionals8h◆Evaluating Causal Reasoning in LLMs: A Case Study in the Challenges of Procedural Generation8h◆AI-Driven Digital Twin Framework for Warehouse Logistics Automation Optimization8h◆SkillCT: A Multidimensional Benchmark and Execution-Free Baseline for Agent Skill Assessment8h◆BridgeRouter: Gain-Aware Instance Routing over Cross-Modal Bridge Portfolios8h◆DecompGym: Benchmarking LLM Agents on Designing Decomposition Algorithms8h◆Inner-Learning: From Hand-Coded Beliefs to Declarative Knowledge in Trick-Taking Games8h◆Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study8h◆Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?8h◆Encode Once, Answer Many: Internalizing Context via Parametric Flow8h◆Masked Erosion: How Acceptance Gates Lose Redundant Safeguards in Self-Improving Agents8h◆Predict the Reward, Not the World: World Reward Models for ML Agents8h◆Value Probes: Learned Value Functions for Beam-Search Red Teaming8h◆From Leniency to Suspicion: How an Open Post-Training Recipe Changes Whom LLM Fact-Checkers Believe8h◆PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation8h◆Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries8h◆MuQ-Eval: An Open-Source Per-Sample Quality Metric for AI Music Generation Evaluation8h◆ThinkJev: One Thought, Every Budget for Calibrated Typed Decisions8h◆What Does an Ontology Catch? Curated, LLM-Generated and Data-Derived Ontologies as Checkers of LLM Agents over Knowledge Graphs8h◆GAUSS: A Workload-Aware Analytical and Monte-Carlo Simulator for LLM Inference Serving8h◆Algorithmic Bias from the Perspectives of Healthcare Professionals8h◆Evaluating Causal Reasoning in LLMs: A Case Study in the Challenges of Procedural Generation8h◆AI-Driven Digital Twin Framework for Warehouse Logistics Automation Optimization8h◆SkillCT: A Multidimensional Benchmark and Execution-Free Baseline for Agent Skill Assessment8h◆BridgeRouter: Gain-Aware Instance Routing over Cross-Modal Bridge Portfolios8h◆DecompGym: Benchmarking LLM Agents on Designing Decomposition Algorithms8h◆Inner-Learning: From Hand-Coded Beliefs to Declarative Knowledge in Trick-Taking Games8h◆Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study8h◆Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?8h◆Encode Once, Answer Many: Internalizing Context via Parametric Flow8h◆Masked Erosion: How Acceptance Gates Lose Redundant Safeguards in Self-Improving Agents8h◆Predict the Reward, Not the World: World Reward Models for ML Agents8h◆Value Probes: Learned Value Functions for Beam-Search Red Teaming8h◆From Leniency to Suspicion: How an Open Post-Training Recipe Changes Whom LLM Fact-Checkers Believe8h◆PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation8h◆Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries8h◆
News/Locating Sycophancy Across Training Stages in Open Language Models
arxiv
PublishedOctober 31, 2026 at 4:00 AM

Locating Sycophancy Across Training Stages in Open Language Models

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

arXiv:2610.37788v1 Announce Type: cross Abstract: Language model assistants sometimes change a correct answer when the user suggests a wrong one. The behavior, called sycophancy, is usually attributed to preference-based post-training, yet early measurements found it in pretrained models. We ask whe

Stay posted· Newsletter

Three short emails a week: Monday, Wednesday and Friday. Top movers, new models and the most-referenced story.

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

Your email is used only to send The Bubble Brief. Privacy policy

Discussion
Source
↗
arxiv
Read original ↗All from arxiv →

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →

Related coverage

More from ARXIV
arxivMuQ-Eval: An Open-Source Per-Sample Quality Metric for AI Music Generation Evaluation8harxivThinkJev: One Thought, Every Budget for Calibrated Typed Decisions8harxivWhat Does an Ontology Catch? Curated, LLM-Generated and Data-Derived Ontologies as Checkers of LLM Agents over Knowledge Graphs8harxivGAUSS: A Workload-Aware Analytical and Monte-Carlo Simulator for LLM Inference Serving8h
The Bubble Brief
MON · WED · FRI

Read AI insights every Monday, Wednesday and Friday — top movers, new models, story of the week.

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

Your email is used only to send The Bubble Brief. Privacy policy

Originally published on arxiv ↗
PrivacyTermsRefundsLegal notice|AboutMethodologyContact

© 2026 Databubble

Built by Marouane Gazouzi · Readyo, AI interview coach

HomeModelsNews