0

LLM Training Puzzles

LLM Training Puzzles by Sasha Rush

Domain
rl-env
License
unknown
Published
Mar 2026

Cite

Notes

Only stored in your browser.

Attribution

Leaderboard scores
prime-hub
Attribution policy →

Top score 62.5% by Gemini 3 Pro Preview - 4 models reporting (2 frontier)

Score history

3
0%25%50%75%100%Aug 25Sep 25Oct 25Nov 25GPT-5Gemini 3 Pro Preview

Top models

4
LLM Training PuzzlesBar chart with 4 bars. Highest value: Gemini 3 Pro Preview at 62.5.
4 models

Related tools

1
View all

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

FAQ

What is LLM Training Puzzles?
LLM Training Puzzles by Sasha Rush
What is the current top score on LLM Training Puzzles?
The top reported score is 62.5% by Gemini 3 Pro Preview, across 4 models reporting (2 from frontier labs).
How can a model improve its LLM Training Puzzles score?
Tools linked to LLM Training Puzzles on Sophon include Training Puzzles RL Env (Prime Community) - RL environments, datasets, and scaffolds that target this eval.
What license is LLM Training Puzzles under?
LLM Training Puzzles is available under unknown.