BlicketTest CausalReasoning
Multi-turn causal reasoning environment where an LLM explores a Blicket-detecting machine to identify which objects activate it under a hidden rule
- Domain
- rl-env
- License
- unknown
- Published
- Feb 2026
Cite
Notes
Only stored in your browser.
Top score 62.0% by Qwen3 30B A3B Instruct 2507 - 3 models reporting (1 frontier)
Score history
2Top models
3
Related tools
1Implementations, trainers, datasets and scaffolds linked to this eval.
FAQ
- What is BlicketTest CausalReasoning?
- Multi-turn causal reasoning environment where an LLM explores a Blicket-detecting machine to identify which objects activate it under a hidden rule
- What is the current top score on BlicketTest CausalReasoning?
- The top reported score is 62.0% by Qwen3 30B A3B Instruct 2507, across 3 models reporting (1 from frontier labs).
- How can a model improve its BlicketTest CausalReasoning score?
- Tools linked to BlicketTest CausalReasoning on Sophon include Blickettest Causalreasoning RL Env (Community) - RL environments, datasets, and scaffolds that target this eval.
- What license is BlicketTest CausalReasoning under?
- BlicketTest CausalReasoning is available under unknown.