Heatmap results may reflect model bias toward certain AI companies
StephenLCasper · x · 2026-07-21
A discussion about interpreting a heatmap result on model/company bias.
The author says the heatmap is the most relevant result they have, and that they did not test interaction hypotheses beyond models with their own companies. Off-diagonal cells do not seem to show anything especially interesting.
In a follow-up thought, they wonder how much of the pattern is driven by pretraining data and “goodness” generalization. One possible explanation is that models may be biased toward Anthropic and against OpenAI/xAI because the web contains more criticism of some companies and more praise of others. They also suggest training dynamics may overall favor companies like Anthropic, though the thought is left unfinished.
Related event: Exploring Why LLMs Show Favoritism Towards Certain AI Companies(2 posts)→
More from Research
- Krea 2 LoKr likeness guide says 750 steps is usually enough for near-perfect face training — LilBrownBebeShoes · 2026-07-22
- PoLar: Dynamically Skipping or Looping LLM Layers for Efficient Inference — ttkciar · 2026-07-22
- Stanford Paper Examines the Institutional Context of AI Benchmarks for Consumers and Regulators — chrmanning · 2026-07-22
- Stanford Team Introduces Gigatoken, the World's Fastest Tokenizer — StanfordAILab · 2026-07-22
- Tabul AI launches Metal TreeSHAP to speed up Shapley values on Apple silicon — Scobleizer · 2026-07-22
- ICML Tutorial: Is Optimization Theory Relevant in 2026? — srush_nlp · 2026-07-22