0

NanochatAquaRat

Frontier

AQuA-RAT multiple-choice reasoning environment aligned with the nanochat RL pipeline

Domain
rl-env
License
unknown
Published
Oct 2025

Cite

Notes

Only stored in your browser.

Attribution

Leaderboard scores
prime-hub
Attribution policy →

Top score 0.0% by Claude 4.1 Opus - 3 models reporting (3 frontier)

Score history

3
0%25%50%75%100%Aug 25Sep 25Claude 4.1 Opus

Top models

3
NanochatAquaRatBar chart with 3 bars. Highest value: GPT-5 Nano at 0.
3 models

Related tools

1
View all

Implementations, trainers, datasets and scaffolds linked to this eval.

FAQ

What is NanochatAquaRat?
AQuA-RAT multiple-choice reasoning environment aligned with the nanochat RL pipeline
What is the current top score on NanochatAquaRat?
The top reported score is 0.0% by Claude 4.1 Opus, across 3 models reporting (3 from frontier labs).
How can a model improve its NanochatAquaRat score?
Tools linked to NanochatAquaRat on Sophon include Nanochataquarat RL Env (Community) - RL environments, datasets, and scaffolds that target this eval.
What license is NanochatAquaRat under?
NanochatAquaRat is available under unknown.