Open-source industrial-grade ASR models supporting Mandarin, Chinese dialects and English, achieving a new SOTA on public Mandarin ASR benchmarks, while also offering outstanding singing lyrics recognition capability.

Curated visual catalog of 155+ vision-language model (VLM/MLLM) architectures: papers, diagrams, training recipes, datasets, and a release timeline for multimodal AI agents.

Official implementation of paper "MiniGPT-5: Interleaved Vision-and-Language Generation via Generative Vokens"

A SOTA Industrial-Grade All-in-One ASR system with ASR, VAD, LID, and Punc modules. FireRedASR2 supports Chinese (Mandarin, 20+ dialects/accents), English, code-switching, and both speech and singing ASR. FireRedVAD supports speech/singing/music in 100+ langs. FireRedLID supports 100+ langs and 20+ zh dialects. FireRedPunc supports zh and en.

Open-source red teaming framework for MLLMs with 42+ attack methods

[ICCV25 Oral] Token Activation Map to Visually Explain Multimodal LLMs

Research Code for Multimodal-Cognition Team in Ant Group

[AAAI 2026] The Official Implementation for "Anomagic: Crossmodal Prompt-driven Zero-shot Anomaly Generation"

🔥 Official code repository for "Unlocking Dense Metric Depth Estimation in VLMs"

[AAAI 2026 Oral] Official repository for InfiGUI-G1. We introduce Adaptive Exploration Policy Optimization (AEPO) to overcome semantic alignment bottlenecks in GUI agents through efficient, guided exploration.

[ACL 2026 Oral] How far can VLMs go in understanding long-horizon action progress from just a single observation? A promising early exploration

[IROS'25 Oral & NeurIPSw'24] Official implementation of "MineDreamer: Learning to Follow Instructions via Chain-of-Imagination for Simulated-World Control "

[ECCV 2024] Official PyTorch Implementation of "How Many Unicorns Are in This Image? A Safety Evaluation Benchmark for Vision LLMs"

OmniAgent (ICML 2026): the first native omni-modal agent for active video perception — a 7B agent that beats Qwen2.5-VL-72B with 73% fewer frames on LVBench.

🦀 Decoder-only LLM built from scratch in pure Rust using Candle — no Python, no PyTorch. Gated DeltaNet + sparse attention, fine-grained MoE, native video/document understanding, long-horizon tool agents, quantization-aware training. Scales: Tiny (25M) to Large (1.3B).

vLLM Qwen 3.6-27B (AWQ-INT4) + DFlash speculative decoding on AMD Strix Halo (gfx1151 iGPU, 128 GB UMA, ROCm 7.13). 24.8 t/s single-stream, vision, tool calling, 256K context, OpenAI-compatible, Docker. Matches DGX Spark FP8+DFlash+MTP at a third of the cost. No CUDA.

Official code for VoLN: Vision-Only Long-Horizon Navigation—Paradigm, Benchmark, and Method. An embodied AI UAV benchmark and VoLN-MLLM agent bridging VLN and multimodal LLMs across 7,210 episodes, AirSim, and real-world flights.

UI-UX: a 4B multimodal LLM for mobile UX defect diagnosis, with UXBench — the first vision–language benchmark for UX reasoning (CVPR 2026 Findings).

The code repository for "Wings: Learning Multimodal LLMs without Text-only Forgetting" [NeurIPS 2024]

Efficient Test-Time Scaling for Small Vision-Language Models, official implementation of the ICLR'26 paper, test-time scaling via test-time augmentation