Stars
Frontier time-series forecasting, zero setup: ci-proto sees 2M+ steps of history and runs on CPU or GPU.
Open time series foundation models for zero-shot probabilistic forecasting
A curated list of papers on Continuous Diffusion and Flow Language Models.
Soft-MTP: Multi-Token Prediction Done Right. KL gradient recovery for faster speculative decoding and better language model training.
Five Bottlenecks, Five Fixes: How We Avoid Leaving Training Performance on the Table
Recipes with Together Post-Training
Executable model integration contracts published on the Hub, run from framework CI
Repository for Timestral, the time-series foundation model by Mistral AI
An RL agent trained only in synthetic markets got better at hedging the real 2020 and 2022 crashes
A scientific testbed for collaboration primitives in heterogeneous LLM-agent swarms
A PostgreSQL extension providing a composite index that efficiently combines vector similarity search with spatial (geometry) filtering.
Unlocking Lossless Speedups in LLMs via Discrete Diffusion
Google DeepMind's WeatherNext 2 at full 0.25° on consumer GPUs. Same model, same results: 6.4 GiB instead of 34 GiB, 7× faster.
Foundation Models for Structured Data
Light-weight infrastructure for express LLM pre-training
A local inference engine for Apple silicon, built around the model.
AI-Driven Scientific, Algorithmic, and Systems Discovery
FineEnvs — RL Environments 101: building and scaling RL environments in the age of LLMs
A general geometry library in LEAN 4.