Stars
OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks
Computer use for about $0.0002 a step: OCR the screen, classify the next action with TypeSafe, click. macOS.
An LLM built without a language model — using TypeSafe Jev's non-generative AI for contextual response selection. Zero hallucination. Zero prompt injection.
Autonomous robotics research with simulation feedback
Value-guided action selection and self-improvement for frozen behaviour-cloning policies
AI agents running research on single-GPU nanochat training automatically
SSH remote workspaces for DeepSeek Harness: parallel projects across hosts, remote file browsing and editing, recursive search, terminals, Git diffs, and workspace-bound Agent tools to edit, build …
Open decision models from 0.8B to 27B: a calibrated probability for every option in one pass, and a short thought only when unsure. Engine for vLLM, llama.cpp and MLX.
JevStyle: local decision models for AI agents. Classify, route and score with a probability per option. systemone-compatible local server, MCP tools, Claude Code guard. Open weights, Apache-2.0.
The NVIDIA-native robotics stack for deploying humanoids, mobile robots and manipulators.
🪢 Open source AI engineering platform: AI Gateway, agent observability, evals, and prompt/skill management, built on OpenTelemetry. One open platform for every AI call and every agent. Self-host or…
Qwen3.8-Flash-Next is the foundation model developed by Qwen Team, Alibaba Group.
SenseNova-U series: Native Unified Paradigm with NEO-unify from the First Principles
EMER, OV-MER (ICML25), AffectGPT (ICML25, Oral), EmoPrefer (ICLR26), AffectGPT-RL (NeurIPS26)
Jev-like family of decision models built on top of Qwen3.5/3.8 you can train and run on your own
Open multimodal decision models, 0.8B to 27B, for text, images and video. Decide with probabilities over your own candidates and locate with bounding boxes, in place of generated text. Free API, ON…
Geometric regulation and correlation dimension tracking for autoregressive text generation in C++
同济大学毕业设计论文 LaTeX 模板 | TongjiThesis - LaTeX Thesis Template for Tongji University
[ICLR 2025] Released code for paper "Spurious Forgetting in Continual Learning of Language Models"
Pretraining and inference code for a large-scale depth-recurrent language model
Embodied intelligence memory for physical AI robots. Built on PowerMem + SeekDB. Zero ROS dependency. Brain-like architecture with spatial-temporal-causal graphs, DTW trajectory search, and gRPC.
Official implementation of the transformer (TF) architecture suggested in a paper entitled "Looped Transformers as Programmable Computers"