PoolBench
Explore score history ↗PoolBench isolates pooling strategies as the experimental variable in concept representation evaluation for decoder-only LLMs. It covers 17 concepts, 19 pooling strategies, and 3 open-weight models (Llama-3.1-8B, Gemma-2-9B, Mistral-7B) on a corpus of 37,693 real-text passages, with primary axis linear separability (AUROC) and diagnostic axes for steering and disentanglement.
- Released
- 2026-05-26
- Evaluates
- General AI, Language & Knowledge, cs.CL
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- unreviewed
PoolBench paper
No reported scores are on record for this benchmark yet.