潜力新星（90天）

An autonomous agentic pipeline that finds, proves, and patches real C memory-safety vulnerabilities end-to-end using a single 7B open-source LLM (Qwen2.5-Coder) on vLLM, with AddressSanitizer as a ground-truth oracle.

10 legit ways to access major AI models (GPT, Claude, Llama, DeepSeek & more) for FREE — step-by-step guides, no subscription, no card.

Benchmarks and notes for running modern LLMs with vLLM on 8x Tesla V100-32GB in 2026.

Cognitive-structured Multimodal Agent (CMA-Harness): a memory-centric agent for long-horizon multimodal understanding, generation, and editing — externalizing visual history into episodic memory with selective retrieval. Includes the M2CA-Bench benchmark.

GLM-5.2 744B with a true 4-bit NVFP4 KV cache on 4x DGX Spark: 42 tok/s peak, 317K-token KV pool (+58.6% vs fp8), needle-verified at 250K depth, serving 316K context

Introducing Muse Glimmer: open-weight 30B agentic multimodal model that runs on your device (Meta). Interactive local agent lab + guide. Apache 2.0 · on-device AI · function calling

让 MinerU 在 AMD 显卡上以 vLLM + hybrid-auto-engine 满血运行！ 本项目是对官方文档中 AMD/ROCm 生态支持的有效补充。针对 MinerU 3.x 提供了一套完整的 ROCm 7.x + PyTorch 2.11.0 + vLLM 源码编译与部署方案，解析质量与速度对标 NVIDIA 旗舰显卡。

AI/ML research project on AMD ROCm

Multimodal document QA: vision + retrieval over PDFs (LLaVA + LlamaIndex)

[CVPR 2026] Flow Matching for Multimodal Distributions

Postgres-Powered Vector Memory for the Agentic Age

Agentic radiology report generator for CT abdomen-pelvis · two vision encoders + MLP head · 220 RATE findings · auditable, deterministic, research-only

Verified multi-source distribution for local LLM weights

chew through any source into clean datasets! a fast ingestion, RAG & fine-tuning toolkit

🍯 自进化 AI 能力研究平台：蜜罐系统 → 沙箱评测引擎 → 训练数据集与 LoRA 微调流水线 / Self-Evolving AI Security Research Platform

Arabic-first expressive zero-shot speech synthesis — Audar-TTS-V1 (Flash + Turbo + Pro). Voice cloning, expression tags, benchmarks, technical report & inference.

Capability-aware vision + paste extension for the pi coding agent. Delegates image analysis only when the active model is text-only; passes through natively for multimodal models (zero delegation).

Honeycomb Lab — hex map + OpenAI gateway control plane for a home AI fleet

The Continuous Verification and Optimization layer for self-hosted LLMs

AI hardware fit calculator for LLM, embedding, reranker, OCR and VLM workloads — VRAM, throughput, licensing and multi-GPU planning