WooooDyy/AgentGym-RL
Framework for training LLM agents on long-horizon, multi-turn decision making with reinforcement learning across a broad set of interactive environments, plus the ScalingInter-RL training method.
Awesome AI Agents › Research Platforms & Simulators
OpenNSWM-Lab/FAROS
FAROS, short for Foundation AutoResearch Operating System, is a Python runtime for automating research workflows. Instead of shipping one fixed AI scientist agent, it treats research automation as a runtime problem and organises execution around four abstractions: Blueprints define the workflow graph, constraints, output contract and validation requirements; Capabilities implement single executable research steps; Profiles bind a blueprint to a concrete execution strategy; and Providers supply the engine behind a capability, whether an LLM, a tool, an API or a human reviewer. The current release candidate ships the first runnable baseline, FAROS-LLM, targeting the LLM research domain. Its blueprint, ml_paper, runs the chain idea to experiment to paper to review across four capabilities: idea_refinement produces an idea session with ranked candidates and a selected candidate, experiment provisions a code project scaffold and experiment record, paper_drafting emits a venue-aware LaTeX project and compiled PDF, and reviewer_simulation returns a structured review with follow-up action items. The runtime lives under backend/app/faros and covers blueprint and profile loading, capability and provider registries, orchestrated execution, event logging, file-backed run, artifact and memory persistence, and baseline verification. Pre-existing domain modules for idea, code, paper, review and platform are reused through capability adapters rather than rewritten. Artifacts are typed, including idea_session, code_project, experiment_record, latex_project, paper_pdf and review_report. The README is explicit about what is not yet present: full DAG scheduling and parallel orchestration, a generalized non-LLM provider ecosystem, a complete experiment execution and evaluation loop, the frontend console, database-backed runtime metadata and a mature cross-domain blueprint library. It requires Python 3.11 or newer and Node 18 or newer, and exposes a FAROS API alongside module-native APIs. The audience is researchers and engineers building configurable automated research pipelines.
https://github.com/OpenNSWM-Lab/FAROS
Also in Research Platforms & Simulators
Purewhiter/mobilegym
MobileGym is a browser-hosted, fully programmable Android-style simulation platform with 28 simulated apps and 416 task templates for benchmarking and reinforcement-learning training of mobile GUI agents.
SimWorld-AI/SimWorld
SimWorld is an Unreal Engine 5 based open-ended simulator for developing and evaluating LLM and VLM agents in realistic physical and social environments such as procedurally generated cities.
Tencent/AICGSecEval
A.S.E is Tencent's repository-level benchmark that measures the security of AI-generated code, building tasks from real GitHub projects and CVE patches and scoring output with static and dynamic analysis.
facebookresearch/meta-agents-research-environments
Meta's research platform for evaluating AI agents in dynamic simulated environments, hosting the Gaia2 benchmark of 800 evolving scenarios across ten universes, with a CLI, a web GUI and a public leaderboard.
HUST-AI-HYZ/MemoryAgentBench
Benchmark and evaluation harness that measures memory in LLM agents across accurate retrieval, test-time learning, long-range understanding and conflict resolution using incremental multi-turn interactions.
DigiRL-agent/digirl
DigiRL is a research project providing code and resources for training autonomous reinforcement learning agents to control Android devices in real-world environments using novel training algorithms and multiple training modes.
xlang-ai/Spider2-V
Spider2-V is a NeurIPS 2024 research project providing a virtual machine environment and dataset to evaluate multimodal agents' ability to automate data science and engineering workflows using state-of-the-art vision-language models.