·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Bringing Value Models Back: Generative Critics for Value Modeling in LLM Reinforcement Learning5h◆Subagents vs Agent Skills: Executing Reusable Knowledge for Long-Horizon Agentic Tasks5h◆Distribution-Consistent Inference for Dynamic Sparse Mixture-of-Experts5h◆In RAG We Trust? Measuring Robustness of Retrieval-Augmented Generation Under Document Poisoning5h◆FrontierChallenge: Evaluating Scientific Workflow Completion5h◆IBIB: A Protocol for Measuring Enterprise AI Systems by Serving Route, Not Model Identifier5h◆AgenticGen: Reward-Guided Agentic Video Generation for Advertising5h◆Strangers to Themselves: What Language Models Say About Themselves Is Generic5h◆Omni Interaction Agent Technical Report5h◆CoGReV: A Confidence-Gated Post-Hoc Non-Monotonic Belief Revision Framework for Phishing Website Classification5h◆Bit-Flip Attacks on Vision-Language-Action Models: Action-Decoding Architecture Shapes the Vulnerability5h◆Can Foundation Models Moderate Online Content? Evaluating Instruction- vs. Example-Driven Policy Operationalization5h◆RiLM: Parameter-Efficient Language Modeling via Geodesic Decoding5h◆Tracing Computation Density in LLMs5h◆Cultural Binding Heads in Language Models5h◆Direct Diversity Optimization for Diverse Successful Trajectories in Preference Post-Training5h◆Left-Branching Transformers Excel at Right-Branching Languages: Data Shapes Word Order Preferences in Language Models5h◆Palmyra x6 Technical Report: An Agentic, Tool-Use Model Post-Trained via Anchored Supervised Fine-Tuning5h◆'Ghaib in Translation' aka Unseen Harm: Measuring Cross-Script Safety Inconsistency with 'Missed-in-Urdu' Scores in LLM Hate Speech Detection5h◆DexterSQL: Deep Schema Exploration and Rule-based Correction for Text-to-SQL Generation5h◆Bringing Value Models Back: Generative Critics for Value Modeling in LLM Reinforcement Learning5h◆Subagents vs Agent Skills: Executing Reusable Knowledge for Long-Horizon Agentic Tasks5h◆Distribution-Consistent Inference for Dynamic Sparse Mixture-of-Experts5h◆In RAG We Trust? Measuring Robustness of Retrieval-Augmented Generation Under Document Poisoning5h◆FrontierChallenge: Evaluating Scientific Workflow Completion5h◆IBIB: A Protocol for Measuring Enterprise AI Systems by Serving Route, Not Model Identifier5h◆AgenticGen: Reward-Guided Agentic Video Generation for Advertising5h◆Strangers to Themselves: What Language Models Say About Themselves Is Generic5h◆Omni Interaction Agent Technical Report5h◆CoGReV: A Confidence-Gated Post-Hoc Non-Monotonic Belief Revision Framework for Phishing Website Classification5h◆Bit-Flip Attacks on Vision-Language-Action Models: Action-Decoding Architecture Shapes the Vulnerability5h◆Can Foundation Models Moderate Online Content? Evaluating Instruction- vs. Example-Driven Policy Operationalization5h◆RiLM: Parameter-Efficient Language Modeling via Geodesic Decoding5h◆Tracing Computation Density in LLMs5h◆Cultural Binding Heads in Language Models5h◆Direct Diversity Optimization for Diverse Successful Trajectories in Preference Post-Training5h◆Left-Branching Transformers Excel at Right-Branching Languages: Data Shapes Word Order Preferences in Language Models5h◆Palmyra x6 Technical Report: An Agentic, Tool-Use Model Post-Trained via Anchored Supervised Fine-Tuning5h◆'Ghaib in Translation' aka Unseen Harm: Measuring Cross-Script Safety Inconsistency with 'Missed-in-Urdu' Scores in LLM Hate Speech Detection5h◆DexterSQL: Deep Schema Exploration and Rule-based Correction for Text-to-SQL Generation5h◆
News/It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO
arxiv
PublishedJune 12, 2026 at 4:00 AM
▼bearish

It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO

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

arXiv:2606.10931v2 Announce Type: replace Abstract: Warning: This paper contains several toxic and offensive statements. Modern large language models (LLMs) are typically aligned through large-scale post-training to ensure fair and reliable behavior. In this work, we investigate how easily such guar

Stay posted· Newsletter

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

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

Discussion
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#bias#safety#language-models#vulnerability

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#bias#safety#language-models#vulnerability

Related coverage

More from ARXIV
arxivBringing Value Models Back: Generative Critics for Value Modeling in LLM Reinforcement Learning5harxivSubagents vs Agent Skills: Executing Reusable Knowledge for Long-Horizon Agentic Tasks5harxivDistribution-Consistent Inference for Dynamic Sparse Mixture-of-Experts5harxivIn RAG We Trust? Measuring Robustness of Retrieval-Augmented Generation Under Document Poisoning5h
The Bubble Brief
WEEKLY

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

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

Originally published on arxiv ↗
HomeModelsNews