Skip to content

v0.1.0

Choose a tag to compare

View all tags
@smirnovlad smirnovlad released this 09 Apr 15:57
· 7 commits to main since this release
v0.1.0
1524cde
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

First PyPI release of ThinkBooster — a framework for test-time compute scaling of LLM reasoning.

Features

  • Strategies: Baseline, Self-Consistency (majority voting), Best-of-N, Beam Search, Adaptive Scaling
  • Backends: vLLM (primary), HuggingFace Transformers, OpenAI-compatible APIs
  • Scorers: Entropy, Perplexity, PRM, Sequence Probability, UHead
  • Evaluation: Exact match with LaTeX/symbolic normalization, LLM judge
  • Datasets: MATH-500, GSM8K, OlympiadBench, GaoKao, Minerva Math, AIME, HumanEval+, MBPP+
  • Extras: Visual debugger for reasoning traces, OpenAI-compatible API endpoint

Install

pip install thinkbooster
Assets 2
Loading