Test Time Compute Python Packages

dual-loop-controller

Hardware-aligned latent deliberation framework decoupling System 2 continuous reasoning from System 1 autoregressive decoding.

verdict

Inference-time scaling for LLMs-as-a-judge.

bestofn

Best-of-N 1.1 — sample a frozen language model N times and select the best answer. 45.3% to 66.5% on GSM8K with a 0.5B model, no training. Eight models measured, all 25,600 trajectories published.

simit

SIMIT: Self-Improving Vision-Language Models via Imagination at Test-Time, official pip library of the paper, SIMIT-ICL: vision-language models that imagine their own in-context demonstrations at test time

fleet-search

Fleet🚀 is an algorithm for distributed test-time scaling of LLM workflows that use local models and can evaluate and rank generated completions.

crucible-ttc

A verifier-guided reasoning engine: generate many reasoning traces from a small open model, score them with a verifier (programmatic checker or a process-reward > model), and **search** — best-of-N → beam → MCTS over reasoning steps — to turn test-time compute into measurable accuracy.

llm-mcts-inference

An experimental project using MCTS to refine LLM responses for better accuracy and decision-making.

hypermcts

🚀 Autonomous Test-Time Compute Scaling & Dynamic KV-Cache Sparsification for Edge & Local LLMs | Pure PyTorch / Ollama MCTS Reasoning Engine (-65% VRAM)

memories-dev

Test-Time Memory Framework: Control Hallucinations in Foundation Models