Havuz

NeoHorse, agent iş akışlarında tool use, coding ve yapılandırılmış kararlar için açık ağırlıklı 4B ve 9B modeller sunuyor.

Neden yükseliyor · 23 günde 1.094 yıldıza ulaşmış; son 28 günde 13 commit ve 6 farklı yazar katkısı da projede yakın dönem geliştirme hareketi olduğunu gösteriyor. README'de model yayını ve arXiv teknik raporu duyurulmuş; ancak ilgi ve ekip aktivitesi, modellerin gerçek kullanımını veya bağımsız başarı teyidini tek başına kanıtlamıyor.

Observal, ekiplerin coding agent bileşenlerini self-hosted bir registry üzerinden yayımlayıp keşfetmesini ve kullanım verilerini incelemesini sağlayan bir platformdur.

Neden yükseliyor · 181 günlük yaşına karşılık 3.533 yıldız, son 28 günde 84 commit ve 8 farklı yazar dikkat çekiyor; bu, yalnızca README iddiasına değil, yakın dönem geliştirme hareketliliğine de işaret ediyor. Bununla birlikte gerçek kurumsal benimseme veya kullanım hacmi verisi sağlanmadığından ilginin üretim kullanımına dönüştüğü doğrulanamıyor.

A clinical decision-evidence benchmark: pull the one fact an answer rests on, and see whether the assistant's action moves with the evidence

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

Minimal and readable coding agent harness implementation in Python to explain the core components of coding agents.

AI Model World, yüzlerce AI modelini fiyat, yetenek, benchmark ve yayın tarihi verileriyle karşılaştıran piksel-art görselli bir web sitesidir.

Neden yükseliyor · Güncel model karşılaştırması ve görsel sunum, hızla değişen AI ekosisteminde ilgi çekici bir kullanım sunuyor. 11 günde 191 yıldız ve son 28 günde 56 commit dikkat çekici bir başlangıç; ancak iki farklı yazar bulunması ve kısa geçmiş, benimsemenin kapsamı ya da sürdürülebilirliği hakkında henüz güçlü kanıt oluşturmuyor.

Open-source autoresearch powered by autonomous coding agents. Run Claude Code, OpenCode, and Codex with grading, shared knowledge, and multi

An intentionally vulnerable OWASP LLM Top 10 training platform for AI Security, Prompt Injection, RAG Security, Agent Security, and GenAI pe

STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization

[NeurIPS 2026 Spotlight] Fast, lossless LLM inference via dual-view diffusion decoding.

Your First LLM-Wiki Conversation Knowledge Base

Jacobian-Brainwash : A manual alignment tool for large language models built on Anthropic's Jacobian Lens. Results are exportable.

Interactive 3D visualization platform for exploring transformer architectures, tensors, and real-time LLM inference.

Molecode presents molecules as code and enables LLMs to operate and reason on chemistry directly.

Transformer architecture explained step by step - the full architecture, every attention variant, positional embeddings, and every layer ins

A recursive self-improvement (RSI) workflow that evolves both target agents (e.g., Pi and DeepSeek Harness) and the RSI system itself.

从 MiniMind 源码读起,再延伸到现代大模型技术体系的中文学习笔记。主线逐行精读预训练 / SFT / DPO / PPO / GRPO 与训练机制;附录 17 篇进阶卷覆盖量化、投机解码、RLHF 全景、模型代际史等 MiniMind 没涉及、但进阶绕不开的主题。

An experiment-driven laboratory for understanding LLM sampling strategies. Explore how Temperature, Top-k, and Top-p influence token probabi

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

An agentic system that auto-optimizes LLM workloads on AMD GPUs.

Simulate AI Personas in Social Feeds with Real-Time Chat Bots 2026

A pure-Python context management layer for LLM systems — retrieval, re-ranking, memory decay, and token-budget enforcement in one pipeline.

211 free visual cards about large language models, from tokenization to deployment. Full collection: 376 cards.

Autonomous AI agent that plays Minecraft on its own — a server-side Fabric mod (MC 1.21.3) where a DeepSeek / OpenAI-compatible LLM drives a

A Large-Scale Knowledge Graph for Automated Scientific Research

Myrtia — agent-native red teaming for LLMs and AI agents: a deliberating council of attacker models (propose, cross-critique, Borda vote, ch

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

Official implementation of “ReSAIL: Mitigating Collapse in Iterative Agent Self-Distillation”

GLM-5.2, a 744 billion parameter mixture of experts model, in a pure C inference engine: quantized to int4, experts streamed from disk, depl

Take the 100 Days of AI challenge - learn one AI concept a day, from machine learning basics to AI agents in production, in 100 days.

Flash weight streaming for MLX: run massive models larger than your RAM on Apple Silicon.

hummingbird is a lightweight, zero dependency runtime for massive open source Mixture of Experts (MoE) language models. It unifies SSD, RAM,

[ICML 2026] When Drafts Evolve: Speculative Decoding Meets Online Learning

Asking for What Was Never Requested: horizontal and vertical proactivity in LLM agents. Need-graph metrics (no LLM judge) and Q&D, which tra

MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems

Benchmark self-evolving Agent upon realistic large-scale file workspaces

AI Systems as Scientific Instruments for Understanding the Mechanisms of Intelligence

TRINITY in Elixir (An Evolved LLM Coordinator): route LLM calls via a small-model hidden-state router + Axon coordination head, with Thinker

🌹 Rose: Range-Of-Slice Equilibration PyTorch optimizer. Stateless optimization through range-normalized gradient updates.

The best model under 200B at near 10 token/s exact batch throughput on a single laptop CPU, with an automatic 8 GB RAM path. Native C, no GP

Run the native 284B-A13B DeepSeek-V4-Flash-0731 LLM locally on one laptop CPU: pure C, 8 GB RAM minimum, no GPU, best TPOT 0.892 s/token, re

Benchmarking data agents for verifiable analytics over heterogeneous workspaces.

Evaan — a fully local, offline-capable AI companion chatbot built with Python & Hugging Face Transformers, running on Qwen2.5-0.5B-Instruct

A full-screen terminal environment for writing fiction with language models.

Qwen3.8-27B on one laptop CPU: up to 2.52 token/s, 8 GB tested, no runtime accuracy loss. Native OpenAI-compatible API with function tools;

Live Benchmark for long Horizon tool using Agentic Systems

Chat with local and cloud AI

EdGameClaw: Convert any learning material into a game-based course — powered by AI.

670 MB/s Swift tokenizer inspired by Marcel Roed’s gigatoken. Zero-copy, NEON/CRC32, WebAssembly, and Embedded Swift.

Tools Are Not Islands: Set-Level Tool Retrieval for LLM Agents via Query-Conditioned Hyperedge Prediction