#bias-mitigation

The latest AI and developer news about #bias-mitigation, with the original source and a short summary.

Arete-27B reduces Chinese government narrative adoption rate from 22% to 0.9% using SFT only

SFT on 52.8 million tokens reduced the Chinese government narrative adoption rate from 22% to 0.9%.

·2·2026.11.13 16:00

Nous Research releases Arete-27B, a post-trained Qwen3.8-27B with reduced state-aligned bias

The model reduces propaganda adoption from 22% to 1% and improves fact scores from 2.32 to 2.98 while maintaining agentic capabilities.

·1·2026.11.07 00:00

Ai2 releases FairStress benchmark revealing asymmetric robustness in LLM decision-making

The 13-million-case benchmark tests 15 models and finds that existing debiasing techniques often fail to fix this identity-based vulnerability without sacrificing accuracy.

·1·2026.10.21 01:00

Google Cloud outlines three-pillar method to audit LLM evaluator biases

The method addresses specific biases like verbosity and position bias, where LLM judges can flip verdicts or miss errors up to 75% of the time.

·1·2026.10.19 09:00

DSO: Direct Steering Optimization for Bias Mitigation

DSO uses RL to optimize steering transformations to reduce bias in VLMs and LLMs.

·2026.04.29 09:00

Best Practices for Deploying Language Models

It presented principles for LLM providers to prevent misuse and document safety.

·1·2022.06.02 21:00