Lists (4)
Starred repositories
SkyRL: A Modular Full-stack RL Library for LLMs
This repository contains comprehensive pricing and configuration data for LLMs. It powers cost attribution for 200+ enterprises running 400B+ tokens through Portkey AI Gateway every day.
A course of learning LLM inference serving on Apple Silicon for systems engineers: build a tiny vLLM + Qwen.
AI-powered terminal assistant for LaTeX academic papers — verifies, fixes, and polishes your paper for conference submission with reviewable diff patches and checkpoint safety.
Tool for generating high quality Synthetic datasets
A command line tool that draw plots on the terminal.
Data generation and training repository for SERA: Soft-Verified Efficient Repository Agents.
Research code artifacts for Code World Model (CWM) including inference tools, reproducibility, and documentation.
The absolute trainer to light up AI agents.
[COLM 2025] "D3: A Dataset for Training Code LMs to Act Diff-by-Diff"
A Claude Code plugin to recover conversation context when Claude loses track
Memorization empirical study validating LoRA-without-regret - Thinking Machines Featured Project
Replication of Open Character Training: fine-tuning LLMs to embody consistent character personas using constitutional methods
Constitutional AI from Base Models - Tinker Featured Project
On-Policy Context Distillation: Comparing distillation methods for knowledge transfer. Teacher-seeded GKD achieves 58-71% GSM8K accuracy while hybrid approaches collapse.
Ideas for projects related to Tinker
An interface library for RL post training with environments.
WHATWG-compliant and fast URL parser written in modern C++, part of Internet Archive, Node.js, Clickhouse, Redpanda, Kong, Telegram, Adguard, Datadog and Cloudflare Workers.
Simple high-throughput inference library
huggingface / yourbench
Forked from sumukshashidhar/yourbench🤗 Benchmark Large Language Models Reliably On Your Data
This is the official code for the paper: Robust Utility-Preserving Text Anonymization Based on Large Language Models
CursorCore: Assist Programming through Aligning Anything
Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels
A repository to unravel the language of GPUs, making their kernel conversations easy to understand
[ICLR 2025] "Training LMs on Synthetic Edit Sequences Improves Code Synthesis" (Piterbarg, Pinto, Fergus)
[ICLR 2025] 🚀 CodeMMLU Evaluator: A framework for evaluating LM models on CodeMMLU MCQs benchmark.