Havuz

Reef, agent isteklerini sunma, feedback toplama, öğrenme döngüsü yürütme ve kabul edilen güncellemeleri sürümleyerek dağıtma işlevlerini bir araya getiren Python altyapısıdır.

Neden yükseliyor · Sürekli öğrenen agent'lar gündemde ve Reef, inference'tan feedback'e, eğitimden sürüm dağıtımına uzanan altyapıyı tek projede toplamayı hedefliyor. Verilen ölçümlerde 28 günde 6.122 yıldız, 362 commit ve 19 farklı yazar güçlü ilgi ve geliştirme faaliyeti gösteriyor; ancak bu veriler tek başına üretim kullanımını veya teknik başarıyı kanıtlamıyor.

MetaClaw, agent konuşmalarından öğrenme sinyalleri çıkarıp skill güncellemeleri ve isteğe bağlı LoRA tabanlı RL eğitimiyle agent’ı zaman içinde iyileştirmeyi amaçlayan bir yazılımdır.

Neden yükseliyor · 3492 yıldız ve README’de duyurulan teknik rapor ile birden fazla agent entegrasyonu dikkat çekmesini açıklayabilir. Ancak son 28 günde commit ve farklı yazar sayısı sıfır; bu nedenle mevcut ivme veya bakım temposu doğrulanamıyor. Teknik raporun HuggingFace Daily Papers sıralaması README’de proje tarafından bildiriliyor.

EvoTrace, Claude Code ve Codex oturumlarını inceleyip eğitim, değerlendirme ve doğrulama için yeniden kullanılabilir varlıklara dönüştürür.

Neden yükseliyor · 39 günde 292 yıldız, coding agent oturumlarından eğitim ve değerlendirme varlığı üretme fikrine ilgi olduğunu gösteriyor; ancak bu ilginin nedeni ölçümlerden anlaşılamıyor. Son 28 günde commit ve farklı yazar olmaması, ilginin devam eden geliştirme veya benimsemeyle desteklendiğine dair kanıt sunmuyor.

JavaScript ile yazılmış, neural network ve genetic algorithm kullanarak sentetik otoyol trafiğinde araç sürüşünü simüle eden etkileşimli bir demo.

Neden yükseliyor · Repo bir günlükken 46 yıldız almış ve canlı demo ile görsel olarak anlaşılır bir konu sunuyor; ancak son 28 günlük commit ve farklı yazar verisi ölçülmediğinden yükselişin devamlılığı veya ekip aktivitesi hakkında kanıt yok.

🔥 On-Policy Self-Distillation in Diffusion Models

[EMNLP 2026] MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research · 浏览器里运行的安卓模拟器 · Browser-hosted A

Halo is an open-source framework built by White Circle for training large language and multimodal models

Turn any repository into verifiable RL environments for coding agents - Harbor tasks you can train on, evaluate and share on the Hugging Fac

FineEnvs — RL Environments 101: building and scaling RL environments in the age of LLMs

《推荐系统：工业架构与核心算法》开源书稿

Creating realistic RL environments requires collaboration between researchers, engineers, and domain experts across many dimensions: artifac

Scikit-learn for Recursive Self-improvement (RSI). A modular framework for recursive harness evolution.

An easy-to-use, fast toolkit to scale up RL post-training on a single node.

verl for a single consumer GPU. PPO, GRPO and on-policy distillation on NVIDIA GPUs.

Your AI intranet: network the computers you already own for inference and training.

Open-source code for training and scaling up energy-based foundation models.

Open-source robotics simulator for coding agents: HTTP/JSON + MCP control, Newton physics, wgpu rendering, ROS 2, and reproducible benchmark

Free Course : Data, Machine Learning, Deep Learning and AI Engineering, the long way round: the math, the code, the systems and the security

NEWTON: Agentic Planning for Physically Grounded Video Generation

Top Dynamic AI World Simulation & Storytelling Tools 2026

ClawCode is claude code inspired implementation in python and Rust focused on agents and experience-based evolution, and is an Open-source c

Self-Tuning Multi-Agent AI Trading System 2026: 8-Source Signal Fusion & Kronos Model

Top 6 AI Paradigms Fighting for Survival in 2026

Source code of paper "RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation"

Dataflow-Oriented Reinforcement Learning for (Multi-)Agentic LLMs

Human-feedback and data-governance infrastructure for RSI research: annotation, review, provenance and frozen dataset candidates.

[arXiv 2026] This is the official PyTorch implementation of "MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators".

An end-to-end framework for Agentic Reinforcement Learning.

RL observability for LLM agents: behavior maps, native trainer metrics, rollout investigations, and Python/SQL analysis.

Your real-life buddy duck: an open-source embodied AI robot duck that walks, hears, sees, and has a life of its own. 让 DuckBuddy 成为你真正的生活伙伴鸭

Self-tuning multi-agent AI trading system. 8-source signal fusion, Bull/Bear/Judge debate on Claude Opus 4.7, Kelly + ATR position sizing. P

Use large-scale AI agent teams to rapidly discover code improvements and "autoresearch" your hardware systems.

The AI study agent that learns how you learn — RL picks the action, FSRS picks the timing, the LLM only writes the quizzes. Chrome extension

Can LLM/MARL agents coordinate in long-horizon, open-ended tasks? NeurIPS 2026 (Evaluations and Datasets).

Reinforcement Learning (RL) environments for the Microban platform, a 30cm open-source and affordable humanoid robot.

🏆Winning Project | ModelGate is a contract-aware AI control plane that ingests customer contracts, extracts SLA/privacy/routing constraints,

Official Repo for Spreadsheet-RL

ICML 2026 benchmark and MechVL models for multimodal understanding of mechanical engineering drawings.

Make your agents learn from experience. One protocol for weights, harness and routing.

A powerful AI-powered chatbot that leverages Neo4j's Cyber Knowledge Graph (CKG) and Groq API to provide intelligent cybersecurity threat in

Reproducible OpenEnv-compatible reinforcement-learning environments for social-media integrity moderation.

🦍 Memorilla: Latent Semantic Memory for LLMs

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

Cooking AGI / AGI 大锅烩 · Bilingual (English / 中文) ML & LLM notes with worked examples, step-by-step figures and coding practice. Free to read

Vela: User Behavior as Supervision Sparks LLM Intelligence — official paper, task templates, and prompts from AllSpark at Xiaohongshu.

View planning with multi-turn VLM agents: ViewSuite 6-DoF benchmark on real ScanNet scenes + iterative RL-SFT training

Beyond Time Shifts (ECCV 2026): a reference-free audio-visual synchronization evaluator built on Qwen2.5-Omni-3B, with SynthSync data, R-GRP

MOS9 open-source humanoid robot hub for RoboCup Kid-Size, connecting hardware, deployment, IPC, policy learning, retargeting, drivers, and s

A programmable, explicit world model on Godot — worlds are pure JSON run by a fixed primitives + interpreter engine. Built by Claude, for Cl

Mental-R1: Aligning LLM Reasoning for Mental Health Assessment