Issues

• Improve external-repository setup and CLI diagnostics

  NVIDIA-NeMo/Gym#4725 · Filter by author anwithk opened on Oct 18, 2026

• Provide supported rollout tagging for external model clients

  NVIDIA-NeMo/Gym#4724 · Filter by author anwithk opened on Oct 18, 2026

• Add run-level totals for model-call capture metrics

  NVIDIA-NeMo/Gym#4723 · Filter by author anwithk opened on Oct 18, 2026

• Expose captured model calls excluded from agent performance totals

  NVIDIA-NeMo/Gym#4722 · Filter by author anwithk opened on Oct 18, 2026

• Fix cached-token double counting in mixed-format usage responses

  NVIDIA-NeMo/Gym#4721 · Filter by author anwithk opened on Oct 18, 2026

• Address Auto-Ontology integration gaps in NeMo Gym

  NVIDIA-NeMo/Gym#4720 · Filter by author anwithk opened on Oct 18, 2026
  1

• Proposal: SECOND STRIKE, a multi-step world-war games environment with a nuclear doomsday clock

  community-request
  NVIDIA-NeMo/Gym#4719 · Filter by author KyleClouthier opened on Oct 18, 2026

• Make gym eval run flags easier to follow for new users

  NVIDIA-NeMo/Gym#4715 · Filter by author cwing-nvidia opened on Oct 17, 2026

• gym dataset commands print a Python traceback for config mistakes instead of a one-line Error:

  community-request
  NVIDIA-NeMo/Gym#4710 · Filter by author sven-rankov opened on Oct 17, 2026

• feat(browsecomp): support runtime-selectable web research providers

  NVIDIA-NeMo/Gym#4709 · Filter by author cwing-nvidia opened on Oct 17, 2026

• vllm_model: two vLLM 0.29 refusal bodies the context-overflow classifier misreads

  NVIDIA-NeMo/Gym#4703 · Filter by author alexandery-nvidia opened on Oct 16, 2026

• docs: refresh stale sample outputs and fix nav, link, and heading hygiene

  documentation
  NVIDIA-NeMo/Gym#4677 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs(reference): fill configuration reference gaps for benchmarks, model servers, and token capture

  documentationmodels
  NVIDIA-NeMo/Gym#4676 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs(training): add a "Train with Gym" overview, post-training eval, and fill training-tutorial gaps

  documentationtraining
  NVIDIA-NeMo/Gym#4675 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs(agents): fix verifiers_agent install guidance and document Harbor / verifiers integrations

  documentation
  NVIDIA-NeMo/Gym#4674 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs(evaluate): add a "Run a benchmark" page and complete the eval workflow

  documentation
  NVIDIA-NeMo/Gym#4673 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs(model-recipes): make the Nemotron 3 Nano recipe copy-paste runnable

  documentationtraining
  NVIDIA-NeMo/Gym#4672 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs: align version pins across install, RL compatibility matrix, and NeMo RL tutorial

  documentationtraining
  NVIDIA-NeMo/Gym#4671 · Filter by author yaoyu-33 opened on Oct 16, 2026

• docs(training): fix offline SFT/DPO rollout conversion snippets

  documentationtraining
  NVIDIA-NeMo/Gym#4670 · Filter by author yaoyu-33 opened on Oct 16, 2026
  

• docs(training): replace stale agent_ref routing with task_source

  documentationtraining
  NVIDIA-NeMo/Gym#4669 · Filter by author yaoyu-33 opened on Oct 16, 2026
  

• docs(evaluate): fix gym eval run examples that fail as written

  CLIdocumentation
  NVIDIA-NeMo/Gym#4668 · Filter by author yaoyu-33 opened on Oct 16, 2026
  

• [epic] docs audit: gym eval and Gym + RL training journeys

  documentation
  NVIDIA-NeMo/Gym#4667 · Filter by author yaoyu-33 opened on Oct 16, 2026

• Coding-agent servers reject generation controls sent by NeMo-RL

  NVIDIA-NeMo/Gym#4656 · Filter by author zyzhou5 opened on Oct 15, 2026

• token-capture: an engine refusal under worker custody reports no truncation verdict

  NVIDIA-NeMo/Gym#4647 · Filter by author alexandery-nvidia opened on Oct 15, 2026

• docs: add Fireworks to RL training

  documentationtraining
  NVIDIA-NeMo/Gym#4645 · Filter by author cwing-nvidia opened on Oct 15, 2026