{"id":51088966,"url":"https://github.com/NVIDIA-NeMo/Gym","last_synced_at":"2026-07-12T04:00:53.459Z","repository":{"id":324434996,"uuid":"1044554840","full_name":"NVIDIA-NeMo/Gym","owner":"NVIDIA-NeMo","description":"Evaluate and improve models and agents using environments","archived":false,"fork":false,"pushed_at":"2026-07-10T21:56:14.000Z","size":70918,"stargazers_count":1037,"open_issues_count":485,"forks_count":224,"subscribers_count":9,"default_branch":"main","last_synced_at":"2026-07-10T22:09:53.606Z","etag":null,"topics":["agents","benchmarks","environments","evaluation","gym","llm","reinforcement-learning","reinforcement-learning-environments","rl-environment","rl-training"],"latest_commit_sha":null,"homepage":"https://docs.nvidia.com/nemo/gym/main/about/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/NVIDIA-NeMo.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":".github/CODEOWNERS","security":"SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-08-25T21:37:55.000Z","updated_at":"2026-07-10T20:14:50.000Z","dependencies_parsed_at":null,"dependency_job_id":"a2843f16-17cf-499b-b57f-c16ae8511ef6","html_url":"https://github.com/NVIDIA-NeMo/Gym","commit_stats":null,"previous_names":["nvidia-nemo/gym"],"tags_count":6,"template":false,"template_full_name":null,"purl":"pkg:github/NVIDIA-NeMo/Gym","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA-NeMo%2FGym","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA-NeMo%2FGym/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA-NeMo%2FGym/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA-NeMo%2FGym/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/NVIDIA-NeMo","download_url":"https://codeload.github.com/NVIDIA-NeMo/Gym/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA-NeMo%2FGym/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35381310,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-12T02:00:06.386Z","response_time":87,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agents","benchmarks","environments","evaluation","gym","llm","reinforcement-learning","reinforcement-learning-environments","rl-environment","rl-training"],"created_at":"2026-06-24T00:00:47.364Z","updated_at":"2026-07-12T04:00:53.449Z","avatar_url":"https://github.com/NVIDIA-NeMo.png","language":"Python","funding_links":[],"categories":["Catalog"],"sub_categories":["Evaluation Harnesses \u0026 Benchmarks"],"readme":"# NeMo Gym\n\n[![PyPI](https://img.shields.io/pypi/v/nemo-gym)](https://pypi.org/project/nemo-gym/)\n[![Python](https://img.shields.io/pypi/pyversions/nemo-gym)](https://pypi.org/project/nemo-gym/)\n[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)\n[![CI](https://github.com/NVIDIA-NeMo/Gym/actions/workflows/unit-tests.yml/badge.svg)](https://github.com/NVIDIA-NeMo/Gym/actions/workflows/unit-tests.yml)\n[![Docs](https://img.shields.io/badge/docs-NVIDIA-brightgreen)](https://docs.nvidia.com/nemo/gym/main/about/)\n\n**[Requirements](#-requirements)** • **[Quick Start](#-quick-start)** • **[Environment Tutorials](#-environment-tutorials)** • **[Available Environments](#-available-environments)** • **[Documentation \u0026 Resources](#-documentation--resources)** • **[Community \u0026 Support](#-community--support)** • **[Citations](#-citations)**\n\nNeMo Gym is a library for evaluating and improving models and agents using environments. NeMo Gym provides infrastructure to develop environments, scalably run evaluation and training, and a collection of popular benchmarks and training environments.\n\nAn environment is the complete system an agent interacts with to complete a task. It consists of a dataset (tasks to solve), an agent harness (how the model interacts with the world), a verifier (task completion scoring), and state (per-task execution context).\n\n## 🎯 When to Use NeMo Gym\n\n- You need to **evaluate models or agents** in stateful environments (e.g. code execution, tool calling, sandboxes)\n- You want **reproducible evaluation** across teams using shared environments and verifiers\n- You need to use environments **at scale** — multiple repeats per task, or thousands of concurrent requests for training\n- You want to **seamlessly transition** between evaluation, agent optimization, and training\n\nIf you're scoring model outputs with a stateless check and don't need scale or training, a script is probably sufficient.\n\n## 🏆 What NeMo Gym Provides\n\n- Modular, extensible interfaces for agents, environments, tasks, and verifiers\n- Environment hub of popular benchmarks and training environments\n- Use your own agents or choose from built-in harnesses\n- Scale to thousands of concurrent environments\n- Train with the RL framework of your choice\n- Battle-tested in production Nemotron training\n\n![NeMo Gym Product Overview](fern/assets/images/product_overview.png)\n\n## 🌎 Ecosystem\n\nNeMo Gym is a component of [NVIDIA NeMo](https://docs.nvidia.com/nemo/gym/main/about/ecosystem#related-nemo-libraries), a GPU-accelerated platform for training generative AI models and optimizing AI agents. NeMo Gym is integrated with the broader agentic ecosystem - see the [Ecosystem](https://docs.nvidia.com/nemo/gym/main/about/ecosystem) page for more details.\n\n**Environment Libraries:** Seamlessly combine environments and benchmarks from other libraries alongside NeMo Gym environments. Examples: \n[Aviary](https://github.com/NVIDIA-NeMo/Gym/tree/main/resources_servers/aviary) • [Harbor](https://github.com/NVIDIA-NeMo/Gym/tree/main/responses_api_agents/harbor_agent) • [OpenEnv](https://github.com/NVIDIA-NeMo/Gym/tree/main/resources_servers/openenv) • [Reasoning Gym](https://github.com/NVIDIA-NeMo/Gym/tree/main/resources_servers/reasoning_gym) • [Verifiers](https://github.com/NVIDIA-NeMo/Gym/tree/main/responses_api_agents/verifiers_agent)\n\n**Training Framework Libraries:** Use environments for SFT and RL training.\n[NeMo RL](https://docs.nvidia.com/nemo/gym/tutorials/training-tutorials/nemo-rl-grpo) • [Unsloth](https://docs.nvidia.com/nemo/gym/tutorials/training-tutorials/unsloth) • [VeRL](https://docs.nvidia.com/nemo/gym/tutorials/training-tutorials)\n\n**Agent Harnesses:** Agent harnesses for evaluation and training available out of the box. Examples:\n[OpenHands](https://github.com/NVIDIA-NeMo/Gym/tree/main/responses_api_agents/swe_agents) • [Mini SWE Agent](https://github.com/NVIDIA-NeMo/Gym/tree/main/responses_api_agents/mini_swe_agent) • [LangGraph](https://github.com/NVIDIA-NeMo/Gym/tree/main/responses_api_agents/langgraph_agent)\n\n\u003e [!IMPORTANT]\n\u003e NeMo Gym is currently in early development. You should expect evolving APIs, incomplete documentation, and occasional bugs. We welcome contributions and feedback - for any changes, please open an issue first to kick off discussion!\n\n## 📣 News\n\n* **[07/01/2026]** [Release v0.4.0](https://github.com/NVIDIA-NeMo/Gym/releases/tag/v0.4.0):\n * **Unified `gym` CLI**: discover agents and benchmarks with `gym list`, catch config errors before a run with `gym env validate`\n * **BLADE diagnostics**: an analysis skill that reads your eval results and reports which tasks failed, why, and the highest-impact fix\n * **Agent skill evaluation**: run the same tasks with different skill sets to measure each skill's impact\n * **Pluggable sandboxes**: run tool-using agents in isolated environments via the sandbox provider of your choice\n * **More agent harnesses**: OpenCode, OpenClaw, and Pi available out of the box\n * **Hosted inference providers**: connect to Fireworks, Together.ai, OpenRouter, and more\n * **New benchmarks**: across science, long-context, and interactive tasks\n\n\u003cdetails\u003e\n\u003csummary\u003ePrevious News\u003c/summary\u003e\n\n* **[06/04/2026]** [Release v0.3.0!](https://github.com/NVIDIA-NeMo/Gym/releases/tag/v0.3.0): 70+ new environments, Nemotron 3 Ultra training datasets, VeRL integration, and out-of-the-box harnesses including Claude Code and Hermes.\n\n\u003c/details\u003e\n\n## 📋 Requirements\n\nNeMo Gym is designed to run on standard development machines:\n\n| Hardware Requirements | Software Requirements |\n| --------------------- | --------------------- |\n| **GPU**: Not required for NeMo Gym library operation\u003cbr\u003e• GPU may be needed for specific resources servers or model inference (see individual server documentation) | **Operating System**:\u003cbr\u003e• Linux (Ubuntu 20.04+, or equivalent)\u003cbr\u003e• macOS (11.0+ for x86_64, 12.0+ for Apple Silicon)\u003cbr\u003e• Windows (via WSL2) |\n| **CPU**: Any modern x86_64 or ARM64 processor (e.g., Intel, AMD, Apple Silicon) | **Python**: 3.12 or higher |\n| **RAM**: Minimum 8 GB (16 GB+ recommended for larger environments) | **Git**: For cloning the repository |\n| **Storage**: Minimum 5 GB free disk space for installation and basic usage | **Internet Connection**: Required for downloading dependencies and API access |\n\n**Additional Requirements**\n\n- **API Keys**: OpenAI API key with available credits (for the quickstart examples)\n - Other model providers supported (Azure OpenAI, self-hosted models via vLLM)\n- **Ray**: Automatically installed as a dependency (no separate setup required)\n\n## 🚀 Quick Start\n\nRequires Python 3.12+ on x86_64 or ARM64 (Linux, macOS, Windows via WSL2). No GPU required. See the [Getting Started](https://docs.nvidia.com/nemo/gym/main/get-started) docs for a more comprehensive walkthrough.\n\n**Install NeMo Gym:**\n\nRequires [uv](https://docs.astral.sh/uv/getting-started/installation/) and Python 3.12+.\n\n```bash\ngit clone git@github.com:NVIDIA-NeMo/Gym.git\ncd Gym\nuv venv --python 3.12 \u0026\u0026 source .venv/bin/activate\nuv sync\n```\n\n**Configure your model:**\n\nThis quickstart uses OpenAI. NeMo Gym supports local and hosted inference — see [Configure Model](https://docs.nvidia.com/nemo/gym/main/model-server) for vLLM, Fireworks, OpenRouter, and others.\n\nCreate `env.yaml` in the project root:\n```yaml\npolicy_base_url: https://api.openai.com/v1\npolicy_api_key: \u003cyour-openai-api-key\u003e\npolicy_model_name: gpt-4.1-2025-04-14\n```\n\n### Run Evaluation\n\nRun your agent on a set of tasks and score the results. This example uses a simple tool calling agent [`simple_agent`](responses_api_agents/simple_agent/README.md) with the [`mcqa`](resources_servers/mcqa/README.md) (multiple-choice Q\u0026A) environment and its included example data.\n\n**1. Start servers**\n\nNeMo Gym uses local servers to coordinate your model, agent, and task verification. Start them first:\n\n```bash\ngym env start \\\n --resources-server mcqa \\\n --model-type openai_model\n```\n\nYou should see three server instances starting:\n\n```text\n[1] mcqa (resources_servers/mcqa)\n[2] mcqa_simple_agent (responses_api_agents/simple_agent)\n[3] policy_model (responses_api_models/openai_model)\n```\n\n**2. Evaluate your agent** \n\nIn a new terminal, run your agent on a single task to verify everything works:\n\n```bash\nsource .venv/bin/activate\n\ngym eval run --no-serve \\\n --agent mcqa_simple_agent \\\n --input resources_servers/mcqa/data/example.jsonl \\\n --output results/mcqa_rollouts.jsonl \\\n --limit 5 \\\n --num-repeats 1\n```\n\nYou should see a progress bar followed by aggregate metrics:\n\n```text\nCollecting rollouts: 100%|██████| 5/5 [01:22\u003c00:00, 16.44s/it]\n\nKey metrics for mcqa_simple_agent:\n{\n \"mean/reward\": 0.8,\n \"pass@1[avg-of-1]/accuracy\": 80.0,\n \"pass@1/accuracy\": 80.0\n}\nFinished rollout collection! View results at:\nFully materialized inputs: results/mcqa_rollouts_materialized_inputs.jsonl\nRollouts: results/mcqa_rollouts.jsonl\nAggregate metrics: results/mcqa_rollouts_aggregate_metrics.json\n```\n\nFor per-task pass rates, see the [`gym eval profile`](https://docs.nvidia.com/nemo/gym/main/reference/cli-commands) command.\n\n### Next Steps\n\n- **[Browse Environments](#-available-environments)** — Browse available environments for evaluation and training.\n- **[Agents](https://docs.nvidia.com/nemo/gym/main/agent-server)** — Explore available agent harnesses and learn how to integrate your own.\n- **[Training](https://docs.nvidia.com/nemo/gym/tutorials/training-tutorials)** — Improve your agent or model with RL or fine-tuning.\n- **[Build Custom Environments](https://docs.nvidia.com/nemo/gym/main/environment-tutorials)** — Create your own evaluation or training environments.\n\n## 🧭 Environment Tutorials\n\nLearn how to build custom environments through hands-on tutorials. Here are popular starting points:\n\n| Name | Demonstrates |\n| ---- | ------------ |\n| [Single Step](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/single-step-environment) | Basic single-step tool calling |\n| [Multi Step](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/multi-step-environment) | Multi-step tool calling |\n| [Session State](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/stateful-environment) | Session state management (in-memory) |\n| [Multi Reward](https://docs.nvidia.com/nemo/gym/main/build-verifiers/multi-reward-verification) | Multiple reward components for evaluation and multi-objective RL (e.g. GDPO) |\n\nSee all [environment tutorials](https://docs.nvidia.com/nemo/gym/main/environment-tutorials) for additional patterns and advanced topics.\n\n## 📦 Available Environments\n\nEnvironments for training and evaluation.\n\nEach resources server includes example data, configuration files, and tests. See each server's README for details.\n\nThe Dataset column links to publicly available datasets (e.g., on HuggingFace). A `-` means the train/validation data has not been publicly released yet, or that it is procedurally generated using a provided script. If no data is released yet, new data can be generated, or the environment can be used as a reference. Each server includes 5 example tasks in `data/example.jsonl`.\n\n\u003c!-- START_TRAINING_SERVERS_TABLE --\u003e\n| Environment | Domain | Description | Value | Train | Validation | License | Config | Dataset |\n| --------------------------------------------- | --------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------- | ----- | ---------- | --------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| Aalcr | other | - | - | - | - | - | \u003ca href='resources_servers/aalcr/configs/aalcr.yaml'\u003eaalcr.yaml\u003c/a\u003e | - |\n| Abstention | rlhf | Train models to abstain when unsure using three-tier reward on HotPotQA with LLM judge | Improve calibration by rewarding abstention over incorrect answers | ✓ | ✓ | Creative Commons Attribution-ShareAlike 4.0 International | \u003ca href='resources_servers/abstention/configs/abstention.yaml'\u003eabstention.yaml\u003c/a\u003e | - |\n| Anyterminal Agent | coding | Terminal Bench run by claude-code natively inside the task container. | Evaluate terminal-task capabilities on Terminal Bench with any Gym agent. | - | - | - | \u003ca href='responses_api_agents/anyterminal_agent/configs/anyterminal_claude_code.yaml'\u003eanyterminal_claude_code.yaml\u003c/a\u003e | - |\n| Anyterminal Agent | coding | Terminal Bench run by the Hermes agent inside the task container. | Evaluate terminal-task capabilities on Terminal Bench with any Gym agent. | - | - | - | \u003ca href='responses_api_agents/anyterminal_agent/configs/anyterminal_hermes.yaml'\u003eanyterminal_hermes.yaml\u003c/a\u003e | - |\n| Arc Agi | knowledge | Solve puzzles designed to test intelligence. See https://arcprize.org/arc-agi. | Improve puzzle-solving capabilities. | - | ✓ | - | \u003ca href='resources_servers/arc_agi/configs/arc_agi.yaml'\u003earc_agi.yaml\u003c/a\u003e | - |\n| Arena Judge | | - | - | - | - | - | \u003ca href='resources_servers/arena_judge/configs/arena_judge.yaml'\u003earena_judge.yaml\u003c/a\u003e | - |\n| Asr With Pc | other | ASR with WER scoring (standard, case-sensitive, punctuation+capitalization) | Improve transcription quality with structural detail | - | - | - | \u003ca href='resources_servers/asr_with_pc/configs/asr_with_pc.yaml'\u003easr_with_pc.yaml\u003c/a\u003e | - |\n| Aviary | agent | Multi-hop question answering on the HotPotQA dataset with Wikipedia search | Improve knowledge and agentic capability | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/aviary/configs/hotpotqa_aviary.yaml'\u003ehotpotqa_aviary.yaml\u003c/a\u003e | - |\n| Aviary | math | GSM8k benchmark with calculator tool | Test math and agentic capability | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/aviary/configs/gsm8k_aviary.yaml'\u003egsm8k_aviary.yaml\u003c/a\u003e | - |\n| Bigcodebench | coding | Verifies model-generated Python solutions against the BigCodeBench unittest suite. | Improve practical, library-rich Python coding capabilities. | - | - | - | \u003ca href='resources_servers/bigcodebench/configs/bigcodebench.yaml'\u003ebigcodebench.yaml\u003c/a\u003e | - |\n| Bird Sql | coding | Text-to-SQL with execution-based evaluation on BIRD dev (1534 SQLite tasks). Binary reward from unordered result-set equality. | Improve text-to-SQL capabilities on BIRD's realistic dev split using execution-based binary reward without an LLM judge. | - | - | - | \u003ca href='resources_servers/bird_sql/configs/bird_sql.yaml'\u003ebird_sql.yaml\u003c/a\u003e | - |\n| Blackjack | games | Blackjack. Model hits or stands. Reward +1 win, 0 draw, -1 loss/bust. | Example gymnasium-style multi-step environment | - | - | - | \u003ca href='resources_servers/blackjack/configs/blackjack.yaml'\u003eblackjack.yaml\u003c/a\u003e | - |\n| Browsecomp Advanced Harness | agent | Model uses search tools to satisfy a user query. | Measure agentic search capability | - | - | - | \u003ca href='resources_servers/browsecomp_advanced_harness/configs/browsecomp_advanced_harness.yaml'\u003ebrowsecomp_advanced_harness.yaml\u003c/a\u003e | - |\n| Bunsenbench Chemistry Mcq | knowledge | Public BunsenBench chemistry multiple-choice benchmark verifier | Measure chemistry MCQ reasoning with source and taxonomy breakdowns | - | - | - | \u003ca href='resources_servers/bunsenbench_chemistry_mcq/configs/bunsenbench_chemistry_mcq.yaml'\u003ebunsenbench_chemistry_mcq.yaml\u003c/a\u003e | - |\n| Calendar | agent | Multi-turn calendar scheduling dataset. User states events and constraints in natural language; model schedules events to satisfy all constraints. | Improve multi-turn instruction following capabilities | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/calendar/configs/calendar.yaml'\u003ecalendar.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-agent-calendar_scheduling'\u003eNemotron-RL-agent-calendar_scheduling\u003c/a\u003e |\n| Calendar | agent | Multi-turn calendar scheduling dataset. User states events and constraints in natural language; model schedules events to satisfy all constraints. | Improve multi-turn instruction following capabilities | ✓ | ✓ | Creative Commons Attribution 4.0 International | \u003ca href='resources_servers/calendar/configs/calendar_v2.yaml'\u003ecalendar_v2.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-Instruction-Following-Calendar-v2'\u003eNemotron-RL-Instruction-Following-Calendar-v2\u003c/a\u003e |\n| Circle Click | other | Click on circles in images | Improve visual grounding and spatial reasoning | - | - | - | \u003ca href='resources_servers/circle_click/configs/circle_click.yaml'\u003ecircle_click.yaml\u003c/a\u003e | - |\n| Circle Count | other | Count circles of a given color in images | Improve visual counting and color recognition | - | - | - | \u003ca href='resources_servers/circle_count/configs/circle_count.yaml'\u003ecircle_count.yaml\u003c/a\u003e | - |\n| Code Fim | coding | Code Fill-in-the-Middle judged by HumanEval-Infilling test suite (single_line, multi_line, random_span, random_span_light) | Improve Python code-infilling capabilities (prefix + completion + suffix) | - | - | - | \u003ca href='resources_servers/code_fim/configs/code_fim.yaml'\u003ecode_fim.yaml\u003c/a\u003e | - |\n| Code Gen | coding | Model must submit the right code to solve a problem | Improve competitive coding capabilities | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/code_gen/configs/code_gen.yaml'\u003ecode_gen.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/nemotron-RL-coding-competitive_coding'\u003enemotron-RL-coding-competitive_coding\u003c/a\u003e |\n| Competitive Coding Challenges | coding | Execution of competitive programming competition questions | Improve competitive coding capabilities on contest-style problems | - | - | - | \u003ca href='resources_servers/competitive_coding_challenges/configs/competitive_coding_challenges.yaml'\u003ecompetitive_coding_challenges.yaml\u003c/a\u003e | - |\n| Critpt | other | Research-level physics problems scored by the Artificial Analysis API | Evaluate model performance on research-level physics reasoning | - | - | - | \u003ca href='resources_servers/critpt/configs/critpt.yaml'\u003ecritpt.yaml\u003c/a\u003e | - |\n| Cvdp | coding | CVDP benchmark dataset for code generation | Evaluate RTL code generation capabilities | - | ✓ | - | \u003ca href='resources_servers/cvdp/configs/cvdp.yaml'\u003ecvdp.yaml\u003c/a\u003e | - |\n| Equivalence Llm Judge | agent | Short bash command generation questions with LLM-as-a-judge | Improve foundational bash and IF capabilities | ✓ | ✓ | GNU General Public License v3.0 | \u003ca href='resources_servers/equivalence_llm_judge/configs/nl2bash-equivalency.yaml'\u003enl2bash-equivalency.yaml\u003c/a\u003e | - |\n| Equivalence Llm Judge | knowledge | Short answer questions with LLM-as-a-judge | Improve knowledge-related benchmarks like GPQA / HLE | - | - | - | \u003ca href='resources_servers/equivalence_llm_judge/configs/equivalence_llm_judge.yaml'\u003eequivalence_llm_judge.yaml\u003c/a\u003e | - |\n| Equivalence Rule | knowledge | Question - Answering with rule-based reward | Improve retrieval and counting capabilities | - | - | - | \u003ca href='resources_servers/equivalence_rule/configs/lc.yaml'\u003elc.yaml\u003c/a\u003e | - |\n| Ether0 | knowledge | ether0 chemistry benchmark verifiers | Evaluate chemistry knowledge and reasoning with ether0 benchmark | - | ✓ | - | \u003ca href='resources_servers/ether0/configs/ether0.yaml'\u003eether0.yaml\u003c/a\u003e | - |\n| Evalplus | coding | Function-completion code judged by EvalPlus base + plus tests (HumanEval+, MBPP+) | Improve Python function-completion capabilities | - | - | - | \u003ca href='resources_servers/evalplus/configs/evalplus.yaml'\u003eevalplus.yaml\u003c/a\u003e | - |\n| Finance Sec Search | agent | SEC EDGAR filing search for financial analysis questions | Enable LLMs to search and analyze SEC filings | - | - | - | \u003ca href='resources_servers/finance_sec_search/configs/finance_sec_search.yaml'\u003efinance_sec_search.yaml\u003c/a\u003e | - |\n| Format Verification | instruction_following | Verify citation/reference markers in model responses via string matching | Improve instruction following for citation format adherence | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/format_verification/configs/citation_format.yaml'\u003ecitation_format.yaml\u003c/a\u003e | - |\n| Format Verification | instruction_following | Verify freeform text formatting (bullets, headings, tables, etc.) via regex patterns | Improve instruction following for text formatting constraints | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/format_verification/configs/freeform_formatting.yaml'\u003efreeform_formatting.yaml\u003c/a\u003e | - |\n| Frontierscience Judge | other | FrontierScience answer grading via single-pass LLM judge | Evaluate FrontierScience Olympiad short answers or Research rubric-scored answers | - | - | - | \u003ca href='resources_servers/frontierscience_judge/configs/frontierscience_judge.yaml'\u003efrontierscience_judge.yaml\u003c/a\u003e | - |\n| Genrm Compare | rlhf | GenRM pairwise comparison for RLHF training | Compare multiple candidate responses using GenRM model | - | - | - | \u003ca href='resources_servers/genrm_compare/configs/genrm_compare.yaml'\u003egenrm_compare.yaml\u003c/a\u003e | - |\n| Google Search | agent | Multi-choice question answering problems with search tools integrated | Improve knowledge-related benchmarks with search tools | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/google_search/configs/google_search.yaml'\u003egoogle_search.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-knowledge-web_search-mcqa'\u003eNemotron-RL-knowledge-web_search-mcqa\u003c/a\u003e |\n| Gpqa Diamond | knowledge | GPQA Diamond multiple-choice question answering problems | Evaluate graduate-level scientific reasoning via MCQ verification | ✓ | - | MIT | \u003ca href='resources_servers/gpqa_diamond/configs/gpqa_diamond.yaml'\u003egpqa_diamond.yaml\u003c/a\u003e | - |\n| Graphwalks | other | Long-context graph-walks (BFS / parents) with F1-over-node-sets grading from openai/graphwalks | Improve long-context multi-step graph reasoning and adjacency-list traversal | - | - | - | \u003ca href='resources_servers/graphwalks/configs/graphwalks.yaml'\u003egraphwalks.yaml\u003c/a\u003e | - |\n| Grl Sokoban | games | Single-box Sokoban in Gymnasium API style. | Model emits one move per turn until the puzzle is solved. | - | - | - | \u003ca href='resources_servers/grl_sokoban/configs/grl_sokoban.yaml'\u003egrl_sokoban.yaml\u003c/a\u003e | - |\n| Grl Tetris | games | Tetris in Gymnasium API style. Model emits one or more moves per turn. | Multi-step Tetris environment | - | - | - | \u003ca href='resources_servers/grl_tetris/configs/grl_tetris.yaml'\u003egrl_tetris.yaml\u003c/a\u003e | - |\n| Gymnasium | other | Base class for Gymnasium-style servers. Not a standalone server. | Reusable base class for step/reset style environments | - | - | - | \u003ca href='resources_servers/gymnasium/configs/gymnasium.yaml'\u003egymnasium.yaml\u003c/a\u003e | - |\n| Harbor Agent | agent | Fast local smoketest task (trivial 1-turn task, no LLM judge) for iterating on the Gym\u003c-\u003eHarbor bridge. | - | ✓ | - | - | \u003ca href='responses_api_agents/harbor_agent/configs/harbor_agent_smoketest_docker.yaml'\u003eharbor_agent_smoketest_docker.yaml\u003c/a\u003e | - |\n| Harbor Agent | agent | Harbor integration for agent harnesses and environments. | Improve models in popular agentic environments supported by Harbor such as Terminus2. | ✓ | - | - | \u003ca href='responses_api_agents/harbor_agent/configs/harbor_agent.yaml'\u003eharbor_agent.yaml\u003c/a\u003e | - |\n| Harbor Agent | agent | Harbor integration for agent harnesses and environments. | Improve models in popular agentic environments supported by Harbor such as Terminus2. | ✓ | - | - | \u003ca href='responses_api_agents/harbor_agent/configs/harbor_agent_daytona.yaml'\u003eharbor_agent_daytona.yaml\u003c/a\u003e | - |\n| Hotpotqa Qa | knowledge | Short-answer QA with deterministic SQuAD-style + alternative-aware substring verification (HotpotQA closed-book). | Improve closed-book multi-hop question-answering accuracy. | - | - | - | \u003ca href='resources_servers/hotpotqa_qa/configs/hotpotqa_qa.yaml'\u003ehotpotqa_qa.yaml\u003c/a\u003e | - |\n| Ifbench | instruction_following | IFBench instruction following evaluation using AllenAI's IFBench library (57 instruction types) | Improve IFBench instruction following | - | - | - | \u003ca href='resources_servers/ifbench/configs/ifbench.yaml'\u003eifbench.yaml\u003c/a\u003e | - |\n| Imo Gradingbench | math | Four-class grading of math proofs — the policy model reads a problem plus a candidate proof and emits one of correct / almost / partial / incorrect as the last word. | Improve the IMO-GradingBench benchmark and proof-grading skill. | - | - | - | \u003ca href='resources_servers/imo_gradingbench/configs/imo_gradingbench.yaml'\u003eimo_gradingbench.yaml\u003c/a\u003e | - |\n| Imo Proofbench Judge | math | IMO ProofBench grader using a strong LLM judge with the IMO 0-7 rubric | Score IMO-style proof submissions with a problem-specific grading rubric | - | - | - | \u003ca href='resources_servers/imo_proofbench_judge/configs/imo_proofbench_judge.yaml'\u003eimo_proofbench_judge.yaml\u003c/a\u003e | - |\n| Indirect Prompt Injection | safety | Indirect prompt injection resistance for multi-domain tool-use agents | Improve agentic security by teaching robustness against tool outputs containing malicious instructions | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/indirect_prompt_injection/configs/indirect_prompt_injection.yaml'\u003eindirect_prompt_injection.yaml\u003c/a\u003e | - |\n| Instruction Following | instruction_following | Instruction following datasets targeting IFEval and IFBench style instruction following capabilities | Improve IFEval and IFBench | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/instruction_following/configs/instruction_following.yaml'\u003einstruction_following.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-instruction_following'\u003eNemotron-RL-instruction_following\u003c/a\u003e |\n| Inverse If | knowledge | Inverse IF instruction-following benchmark with per-task LLM judge | - | ✓ | - | TBD | \u003ca href='resources_servers/inverse_if/configs/inverse_if.yaml'\u003einverse_if.yaml\u003c/a\u003e | - |\n| Jailbreak Detection | safety | Jailbreak detection with Nemotron judge + combined reward | Improve Jailbreak Robustness and Safety/Security Behavior Guide Enforcement | - | - | - | \u003ca href='resources_servers/jailbreak_detection/configs/jailbreak_detection_nemotron_combined_reward_tp8.yaml'\u003ejailbreak_detection_nemotron_combined_reward_tp8.yaml\u003c/a\u003e | - |\n| Labbench2 Vlm | knowledge | labbench2 VLM benchmarks: scientific figure/table QA (figqa2, tableqa2), protocol troubleshooting (protocolqa2), LLM-as-judge | Measure scientific reasoning on figures, tables, and lab protocols | - | ✓ | - | \u003ca href='resources_servers/labbench2_vlm/configs/labbench2_vlm.yaml'\u003elabbench2_vlm.yaml\u003c/a\u003e | - |\n| Longmt Eval | other | Document-level MT verifier for pg19 books using the SEGALE pipeline (ersatz segment → LASER2 embed → vecalign align → COMETKiwi score) | Rewards long-form book translation at the document level using reference-free COMETKiwi scores as the RL reward signal. | - | - | - | \u003ca href='resources_servers/longmt_eval/configs/longmt_pg19.yaml'\u003elongmt_pg19.yaml\u003c/a\u003e | - |\n| Longmt Eval | other | Document-level MT verifier for wmt24pp short docs using the SEGALE pipeline (ersatz segment → LASER2 embed → vecalign align → COMETKiwi score). | Rewards document-level translation quality across 55 language pairs using reference-free COMETKiwi scores as the RL reward signal. | - | - | - | \u003ca href='resources_servers/longmt_eval/configs/longmt_wmt24pp.yaml'\u003elongmt_wmt24pp.yaml\u003c/a\u003e | - |\n| Longmt Eval | other | Document-level MT verifier using the SEGALE pipeline (ersatz segment → LASER2 embed → vecalign align → COMETKiwi score) | Rewards long-form translation quality at the document level using reference-free COMETKiwi scores as the RL reward signal. | - | - | - | \u003ca href='resources_servers/longmt_eval/configs/longmt_eval.yaml'\u003elongmt_eval.yaml\u003c/a\u003e | - |\n| Math Advanced Calculations | agent | An instruction following math environment with counter-intuitive calculators | Improve instruction following capabilities in specific math environments | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/math_advanced_calculations/configs/math_advanced_calculations.yaml'\u003emath_advanced_calculations.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-math-advanced_calculations'\u003eNemotron-RL-math-advanced_calculations\u003c/a\u003e |\n| Math Formal Lean | math | Lean4 formal proof verification environment | Improve formal theorem proving capabilities | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/math_formal_lean/configs/nemotron_clean_easy.yaml'\u003enemotron_clean_easy.yaml\u003c/a\u003e | - |\n| Math Formal Lean | math | Lean4 formal proof verification environment | Improve formal theorem proving capabilities | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/math_formal_lean/configs/nemotron_first_try_hard.yaml'\u003enemotron_first_try_hard.yaml\u003c/a\u003e | - |\n| Math Formal Lean | math | Lean4 formal proof verification environment | Improve formal theorem proving capabilities | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/math_formal_lean/configs/nemotron_medium_500.yaml'\u003enemotron_medium_500.yaml\u003c/a\u003e | - |\n| Math Formal Lean | math | Lean4 formal proof verification environment | Improve formal theorem proving capabilities | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/math_formal_lean/configs/nemotron_very_easy.yaml'\u003enemotron_very_easy.yaml\u003c/a\u003e | - |\n| Math Formal Lean | math | Lean4 formal proof verification environment | Improve formal theorem proving capabilities | ✓ | - | MIT | \u003ca href='resources_servers/math_formal_lean/configs/math_formal_lean.yaml'\u003emath_formal_lean.yaml\u003c/a\u003e | - |\n| Math Formal Lean | math | Lean4 formal proof verification environment with multi-turn self-correction | Improve formal theorem proving capabilities | ✓ | - | MIT | \u003ca href='resources_servers/math_formal_lean/configs/math_formal_lean_multi_turn.yaml'\u003emath_formal_lean_multi_turn.yaml\u003c/a\u003e | - |\n| Math Proof Judgement | math | Binary judgement of math proofs — the policy model reads a problem plus a candidate proof and outputs Judgement: Yes/No. | Improve the NVIDIA ProofBench judge benchmark and math-proof verification skill. | - | - | - | \u003ca href='resources_servers/math_proof_judgement/configs/math_proof_judgement.yaml'\u003emath_proof_judgement.yaml\u003c/a\u003e | - |\n| Math With Autograder | math | Math QA verified by a Skills-style autograder LLM judge with math-verify symbolic fallback | Score hard-math benchmarks (e.g. IMO AnswerBench) where the judge is a unidirectional Correct/Incorrect grader | - | - | - | \u003ca href='resources_servers/math_with_autograder/configs/math_with_autograder.yaml'\u003emath_with_autograder.yaml\u003c/a\u003e | - |\n| Math With Code | math | Model solves competitive math problems using simple calculator tools | Improve math and simple tool use capabilities | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/math_with_code/configs/math_with_code.yaml'\u003emath_with_code.yaml\u003c/a\u003e | - |\n| Math With Judge | math | DAPO17k math dataset with math-verify | Improve math capabilities including AIME 24 / 25 | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/math_with_judge/configs/dapo17k.yaml'\u003edapo17k.yaml\u003c/a\u003e | - |\n| Math With Judge | math | Hermes Agent with terminal, file, code_execution, skills, todo toolsets on OpenMathReasoning math dataset with math-verify and LLM-as-a-judge | Improve model math capabilities in hermes agent harness such as AIME25 | ✓ | - | Creative Commons Attribution 4.0 International | \u003ca href='resources_servers/math_with_judge/configs/math_with_judge_hermes_agent.yaml'\u003emath_with_judge_hermes_agent.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-math-OpenMathReasoning'\u003eNemotron-RL-math-OpenMathReasoning\u003c/a\u003e |\n| Math With Judge | math | MathStackOverflow math dataset with math-verify | Improve math capabilities including AIME 24 / 25 | ✓ | ✓ | Creative Commons Attribution-ShareAlike 4.0 International | \u003ca href='resources_servers/math_with_judge/configs/math_stack_overflow.yaml'\u003emath_stack_overflow.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-math-stack_overflow'\u003eNemotron-RL-math-stack_overflow\u003c/a\u003e |\n| Math With Judge | math | OpenClaw agent harness on OpenMathReasoning math dataset with math-verify | Improve model math capabilities in the OpenClaw agent harness | - | - | - | \u003ca href='resources_servers/math_with_judge/configs/math_with_judge_openclaw_agent.yaml'\u003emath_with_judge_openclaw_agent.yaml\u003c/a\u003e | - |\n| Math With Judge | math | OpenCode agent harness on OpenMathReasoning math dataset with math-verify | Improve model math capabilities in the OpenCode agent harness | - | - | - | \u003ca href='resources_servers/math_with_judge/configs/math_with_judge_opencode_agent.yaml'\u003emath_with_judge_opencode_agent.yaml\u003c/a\u003e | - |\n| Math With Judge | math | OpenMathReasoning math dataset with math-verify and LLM-as-a-judge | Improve math capabilities including AIME 24 / 25 | ✓ | ✓ | Creative Commons Attribution 4.0 International | \u003ca href='resources_servers/math_with_judge/configs/math_with_judge.yaml'\u003emath_with_judge.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-math-OpenMathReasoning'\u003eNemotron-RL-math-OpenMathReasoning\u003c/a\u003e |\n| Math With Judge | math | pi agent harness on OpenMathReasoning math dataset with math-verify | Improve model math capabilities in the pi agent harness | - | - | - | \u003ca href='resources_servers/math_with_judge/configs/math_with_judge_pi_agent.yaml'\u003emath_with_judge_pi_agent.yaml\u003c/a\u003e | - |\n| Mcqa | knowledge | Multi-choice question answering problems | Improve benchmarks like MMLU / GPQA / HLE | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/mcqa/configs/mcqa.yaml'\u003emcqa.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-knowledge-mcqa'\u003eNemotron-RL-knowledge-mcqa\u003c/a\u003e |\n| Mini Swe Agent | coding | Software engineering tasks driven by mini-swe agent harness. | Improve agentic software engineering capabilities. | ✓ | ✓ | MIT | \u003ca href='responses_api_agents/mini_swe_agent/configs/mini_swe_agent.yaml'\u003emini_swe_agent.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/SWE-Gym/SWE-Gym'\u003eSWE-Gym\u003c/a\u003e |\n| Mrcr | other | Multi-round coreference resolution over multi-turn conversations with prefix-gated SequenceMatcher grading | Improve long-context in-context retrieval and needle-count-aware reasoning | - | - | - | \u003ca href='resources_servers/mrcr/configs/mrcr.yaml'\u003emrcr.yaml\u003c/a\u003e | - |\n| Multichallenge | knowledge | Targets inference memory, instruction retention, version editing, and self-coherence. | Improve complex multi-turn conversational capability | ✓ | - | Creative Commons Attribution 4.0 International | \u003ca href='resources_servers/multichallenge/configs/multichallenge_nrl.yaml'\u003emultichallenge_nrl.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1'\u003eNemotron-RL-Instruction-Following-MultiTurnChat-v1\u003c/a\u003e |\n| Newton Bench | math | Scientific law discovery tasks through agentic experimentation across 12 physics domains | Improve science, reasoning, and tool use capabilities | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/newton_bench/configs/newton_bench.yaml'\u003enewton_bench.yaml\u003c/a\u003e | - |\n| Ns Tools | agent | NeMo Skills tool execution with math verification | - | - | - | - | \u003ca href='resources_servers/ns_tools/configs/ns_tools.yaml'\u003ens_tools.yaml\u003c/a\u003e | - |\n| Nvarc | knowledge | ARC-AGI inductive mode: model outputs Python code with transform() | Improve ARC-AGI puzzle-solving by inducing executable transformation programs | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/nvarc/configs/inductive.yaml'\u003einductive.yaml\u003c/a\u003e | - |\n| Nvarc | knowledge | ARC-AGI transductive mode: model outputs grid directly | Improve ARC-AGI puzzle-solving by directly predicting transformed grids | ✓ | ✓ | Apache 2.0 | \u003ca href='resources_servers/nvarc/configs/transductive.yaml'\u003etransductive.yaml\u003c/a\u003e | - |\n| Omniscience | knowledge | Omniscience factual knowledge QA with LLM judge verification | Evaluate factual recall and calibration via LLM-graded open-ended QA | - | - | - | \u003ca href='resources_servers/omniscience/configs/omniscience.yaml'\u003eomniscience.yaml\u003c/a\u003e | - |\n| Openenv | agent | Echo environment via OpenEnv (MCP). Echoes messages back with length-based rewards. | - | - | - | - | \u003ca href='resources_servers/openenv/configs/openenv_echo.yaml'\u003eopenenv_echo.yaml\u003c/a\u003e | - |\n| Openenv | coding | Python code execution environment via OpenEnv. Executes code and returns stdout/stderr. | - | - | - | - | \u003ca href='resources_servers/openenv/configs/openenv_coding.yaml'\u003eopenenv_coding.yaml\u003c/a\u003e | - |\n| Openenv | games | Maze navigation environment via OpenEnv. Agent navigates an 8x8 grid to find the exit. | - | - | - | - | \u003ca href='resources_servers/openenv/configs/openenv_maze.yaml'\u003eopenenv_maze.yaml\u003c/a\u003e | - |\n| Over Refusal Detection | | - | - | ✓ | - | TBD | \u003ca href='resources_servers/over_refusal_detection/configs/over_refusal_detection.yaml'\u003eover_refusal_detection.yaml\u003c/a\u003e | - |\n| Physics Judge | math | Physics QA verified by NeMo Skills' physics judge LLM with math-verify symbolic fallback | Score open-ended physics benchmarks (e.g. PHYSICS) where the judge emits [Correct] / [Incorrect] verdicts | - | - | - | \u003ca href='resources_servers/physics_judge/configs/physics_judge.yaml'\u003ephysics_judge.yaml\u003c/a\u003e | - |\n| Polymath | math | PolyMath multilingual math benchmark with weighted (difficulty) and per-language metrics | Improve multilingual math reasoning across 18 languages and 4 difficulty tiers | - | - | - | \u003ca href='resources_servers/polymath/configs/polymath.yaml'\u003epolymath.yaml\u003c/a\u003e | - |\n| Proof Genselect | math | Pairwise proof selection with binary correctness reward | - | - | - | - | \u003ca href='resources_servers/proof_genselect/configs/proof_genselect.yaml'\u003eproof_genselect.yaml\u003c/a\u003e | - |\n| Proof Judge | math | Theorem proving with verifier + meta-verifier judge (combined env) | - | - | - | - | \u003ca href='resources_servers/proof_judge/configs/proof_judge.yaml'\u003eproof_judge.yaml\u003c/a\u003e | - |\n| Proof Verification | math | Proof verification scored against ground truth and meta-verifier agreement | - | - | - | - | \u003ca href='resources_servers/proof_verification/configs/proof_verification.yaml'\u003eproof_verification.yaml\u003c/a\u003e | - |\n| Rdkit Chemistry | knowledge | Molecular chemistry question answering: calculate properties of SMILES. Includes a mix of tool-use (python + rdkit) and no-tool-use questions. | Improve molecular reasoning and SMILES parsing. | ✓ | - | TBD | \u003ca href='resources_servers/rdkit_chemistry/configs/rdkit_chemistry.yaml'\u003erdkit_chemistry.yaml\u003c/a\u003e | - |\n| Reasoning Gym | knowledge | Claude Code agent harness for reasoning gym tasks | Evaluate model capabilities in the Claude Code agent harness | ✓ | - | Creative Commons Attribution 4.0 International | \u003ca href='resources_servers/reasoning_gym/configs/reasoning_gym_claude_code_agent.yaml'\u003ereasoning_gym_claude_code_agent.yaml\u003c/a\u003e | \u003ca href='https://huggingface.co/datasets/nvidia/Nemotron-RL-ReasoningGym-v1'\u003eNemotron-RL-ReasoningGym-v1\u003c/a\u003e |\n| Reasoning Gym | knowledge | Claude Code agent harness for reasoning gym tasks, via a Gym model server's /v1/messages | Showcase Claude Code running against any Gym model backend | - | - | - | \u003ca href='resources_servers/reasoning_gym/configs/reasoning_gym_claude_code_agent_model_server.yaml'\u003ereasoning_gym_claude_code_agent_model_server.yaml\u003c/a\u003e | - |\n| Reasoning Gym | knowledge | LangGraph orchestrator agent compatible with resource servers that do not use tools; enables diverse agent training data and test time scaling vs a simple agent, extensible to use tools or other agent architectures | Iterative test time scaling for improved performance in reasoning tasks | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/reasoning_gym/configs/orchestrator_agent.yaml'\u003eorchestrator_agent.yaml\u003c/a\u003e | - |\n| Reasoning Gym | knowledge | LangGraph parallel thinking agent compatible with resource servers that do not use tools; enables diverse agent training data and test time scaling vs a simple agent, extensible to use tools or other agent architectures | Iterative test time scaling for improved performance in reasoning tasks | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/reasoning_gym/configs/parallel_thinking_agent.yaml'\u003eparallel_thinking_agent.yaml\u003c/a\u003e | - |\n| Reasoning Gym | knowledge | LangGraph reflection agent compatible with resource servers that do not use tools; provides iterative reflection for diverse agent training data and test time scaling, extensible to use tools or other agent architectures | Iterative test time scaling for improved performance in reasoning tasks | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/reasoning_gym/configs/reflection_agent.yaml'\u003ereflection_agent.yaml\u003c/a\u003e | - |\n| Reasoning Gym | knowledge | LangGraph ReWOO agent compatible with resource servers that do not use tools; enables diverse agent training data and test time scaling vs a simple agent, extensible to use tools or other agent architectures | Iterative test time scaling for improved performance in reasoning tasks | ✓ | - | Apache 2.0 | \u003ca href='resources_servers/reasoning_gym/configs/rewoo_agent.yaml'\u003erewoo_agent.yaml\u003c/a\u003e | - |\n| Reasoning Gym | knowledge | Over 100 tasks including algebra, arithmetic, computation, cognition, geometry, graph theory, logic, and many common games. | Improve robustness, generalization, broad knowledge and reasoning | ✓ | - | Creative Commons Attribution 4.0 International | \u003ca href='resources_servers/reasoning_gym/configs/reasoning_gym.yaml'\u003ereasoning_gym.ya","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FNVIDIA-NeMo%2FGym","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FNVIDIA-NeMo%2FGym","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FNVIDIA-NeMo%2FGym/lists"}