·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Beyond a Single Direction: Chain-of-Thought Disrupts Simple Steering of Refusal4h◆RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching4h◆VarRate: Training-Free Variable-Rate KV Cache Compression for Long-Context LLMs4h◆Adaptive Multi-Step Lookahead Decoding for Diffusion Language Models4h◆An Auto-Scaling Approach for Serverless Environments Based on a Multi-Expert Consensus Mechanism4h◆Cache-Aware Prompt Compression:A Two-Tier Cost Model for LLM API Caching4h◆Candidate Attended Dialogue State Tracking Using BERT4h◆An Exam for Active Observers4h◆A Neuro-Symbolic Approach for Probabilistic Reasoning on Graph Data4h◆Agent Step Value: Auditing Evaluator-Channel Reversals in Black-Box Agent Traces4h◆Evidence-Aware MapReduce for Forkable Compute4h◆From Plausible to Actionable: A Position on LLM Self-Explanations4h◆Robust Explanations for User Trust in Enterprise NLP Systems4h◆Brain-CLIPLM: Semantic Compression for EEG-to-Text Decoding4h◆Before the Action: Benchmarking LLMs on Prospective Hypothesis Discovery4h◆Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers4h◆SODA: Semi On-Policy Black-Box Distillation for Large Language Models4h◆Graph Coloring Approach to Solving Sudoku with Oscillatory Neural Networks4h◆MLLM-DataEngine: Closing the Loop of Multimodal Instruction Tuning Data Generation4h◆Retraining Seeks Stable Signals4h◆Beyond a Single Direction: Chain-of-Thought Disrupts Simple Steering of Refusal4h◆RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching4h◆VarRate: Training-Free Variable-Rate KV Cache Compression for Long-Context LLMs4h◆Adaptive Multi-Step Lookahead Decoding for Diffusion Language Models4h◆An Auto-Scaling Approach for Serverless Environments Based on a Multi-Expert Consensus Mechanism4h◆Cache-Aware Prompt Compression:A Two-Tier Cost Model for LLM API Caching4h◆Candidate Attended Dialogue State Tracking Using BERT4h◆An Exam for Active Observers4h◆A Neuro-Symbolic Approach for Probabilistic Reasoning on Graph Data4h◆Agent Step Value: Auditing Evaluator-Channel Reversals in Black-Box Agent Traces4h◆Evidence-Aware MapReduce for Forkable Compute4h◆From Plausible to Actionable: A Position on LLM Self-Explanations4h◆Robust Explanations for User Trust in Enterprise NLP Systems4h◆Brain-CLIPLM: Semantic Compression for EEG-to-Text Decoding4h◆Before the Action: Benchmarking LLMs on Prospective Hypothesis Discovery4h◆Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers4h◆SODA: Semi On-Policy Black-Box Distillation for Large Language Models4h◆Graph Coloring Approach to Solving Sudoku with Oscillatory Neural Networks4h◆MLLM-DataEngine: Closing the Loop of Multimodal Instruction Tuning Data Generation4h◆Retraining Seeks Stable Signals4h◆
News/Sycophancy in GPT-4o: what happened and what we’re doing about it
openai
PublishedApril 29, 2025 at 6:00 PM

Sycophancy in GPT-4o: what happened and what we’re doing about it

Source
openai.comfull article ↗
Read on openai→
Publisher summary· verbatim

We have rolled back last week’s GPT‑4o update in ChatGPT so people are now using an earlier version with more balanced behavior. The update we removed was overly flattering or agreeable—often described as sycophantic.

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
↗
openai
Read original ↗All from openai →

No replies yet. Be first.

Source
↗
openai
Read original ↗All from openai →
The Bubble Brief
WEEKLY

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

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

Originally published on openai ↗
HomeModelsNews