Dev
GitHub repos gaining traction - what high-signal users are starring and what's climbing the board, captured daily and enriched from GitHub. Raw material for spotting new tech and patterns worth building on.
653
repos tracked
153
surfaced this week
141
created < 30d
Python
top language
25 repos
-
slime is an LLM post-training framework for RL Scaling.
-
Agentic RL Training at Scale
-
Training open models for agentic phone use with real-app and mock-app environments.
-
An interface library for RL post training with environments.
-
Ongoing research training transformer models at scale
-
ABC: Scalable Behavior Cloning with Open Data, Training, and Evaluation
-
An LLM post-training framework with vLLM for RL Scaling
-
Experimental fork of Karpathy's nanoGPT modifying the architecture and training loop to make per-source influence over predictions identifiable and controllable.
-
Arctic Training and Inference Platform
-
AllenAI's post-training codebase
-
Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.
-
[ICML '26] Code repo for the paper entitled "Convex Dataset Valuation for Post-Training" at ICML 2026.
-
A PyTorch native platform for training generative AI models
-
Lens is a 3.8B-parameter text-to-image diffusion model that achieves quality competitive with and in several cases surpassing models like FLUX and SD3, while requiring significantly less training compute. Key ideas include maximizing data information density per batch and accelerating convergence.
-
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
-
Predictive Style Matching (PSM), is method in which an offline predictor maps the robot’s lower-body state history and velocity commands to interpretable upper-body joint and gait targets that shape the rewards during training.
-
Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.
-
★ 2 mvanhorn/unslothUnsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.
-
Post-training with Tinker
-
[CVPR 2026 Oral] "INSID3: Training-Free In-Context Segmentation with DINOv3"
-
🚀 Ultra Recipe for Training Long-Horizon Search Agents - matching frontier AI's search capability with a 20B model + stateful harness
-
nnScaler: Compiling DNN models for Parallel Training
-
Post-training framework for large models, from new objectives to new rollout systems.
-
Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.
-
Compact and Agent-Native MoE Training System