All pull requests

Pull requests

• [GDPVal] Coerce a rendered paths list nested inside the finish argument

  area:agentbugcomplexity:lowneeds-review
  #3659·agronskiy opened Sep 23, 2026·

  ·

• feat(orchestration): express prefill/decode disaggregation

  area:orchestrationcomplexity:highfeatureneeds-review
  #3657·wprazuch opened Sep 23, 2026·

  ·

• feat(orchestration): let a benchmark run its own harness

  area:orchestrationcomplexity:lowfeatureneeds-review
  #3656·wprazuch opened Sep 23, 2026·

  ·

• fix(orchestration): write job.sh owner-only

  area:orchestrationbugcomplexity:lowneeds-review
  #3655·wprazuch opened Sep 23, 2026·

  ·

• feat(orchestration): let a service claim its own nodes

  area:orchestrationcomplexity:highfeatureneeds-review
  #3654·wprazuch opened Sep 23, 2026·

  ·

• design(environments): prototype Harbor task authoring

  #3653·cwing-nvidia opened Sep 23, 2026·

  ·
  1

• fix(sandbox): stop local subprocesses on cancellation

  #3652·yaoyu-33 opened Sep 23, 2026·

  ·
  1

• feat(agentdyn): add AgentDyn indirect prompt-injection benchmark

  area:environmentcommunity-requestcomplexity:highfeatureneeds-review
  #3651·reinainblood opened Sep 23, 2026·

  ·
  1

• feat(exploitbench): add ExploitBench V8 capability-ladder benchmark

  area:environmentcommunity-requestcomplexity:highfeatureneeds-review
  #3650·reinainblood opened Sep 23, 2026·

  ·
  1

• fix(gdpval): don't count negative rubric points toward max_possible

  area:environmentbugcomplexity:lowneeds-review
  #3648·cmunley1 opened Sep 23, 2026·

  ·
  2

• feat(launcher): add Lightning fast Terminal-Bench 2.1 profile for the pool agent

  community-request
  #3647·brianfitzgerald opened Sep 23, 2026·

  ·
  1

• feat(token-capture): add direct ledger manifest reader

  area:trainingfeature
  #3646·pthombre opened Sep 23, 2026·

  ·
  1

• feat(benchmarks): add Indic MATH-500

  area:environmentcommunity-requestcomplexity:mediumfeatureneeds-review
  #3645·anushaknvidia opened Sep 23, 2026·

  ·
  1

• feat(indic mbpp): add Indic MBPP

  area:environmentcommunity-requestcomplexity:mediumfeatureneeds-review
  #3644·anushaknvidia opened Sep 23, 2026·

  ·
  1

• [ApexAgents] Relay policy traffic over a unix socket for network-isolated sandboxes

  area:agentcomplexity:mediumfeatureneeds-review
  #3642·Kh4L opened Sep 23, 2026·

  ·
  1

• fix(sandbox): fail fast when attaching to a missing OpenSandbox PTY session

  area:sandboxbugcomplexity:lowready-to-merge
  #3641·hemildesai opened Sep 23, 2026·

  ·
  3

• feat(agent): run Codex inside native task sandboxes

  area:agentfeature
  #3640·yaoyu-33 opened Sep 22, 2026·

  ·
  1

• feat(agent): support native OpenCode in its canonical package

  area:agentfeature
  #3639·yaoyu-33 opened Sep 22, 2026·

  ·
  2

• feat(telemetry): count finished rollouts by outcome, class and reason

  area:core
  #3638·iguehring-nvidia opened Sep 22, 2026·

  ·
  2

  

• feat(agent): run OpenClaw inside native task sandboxes

  area:agentfeature
  #3637·yaoyu-33 opened Sep 22, 2026·

  ·
  1

• Serve SEC search from a local corpus in both finance servers

  area:environmentcomplexity:highfeatureneeds-review
  #3636·ushnish-de opened Sep 22, 2026·

  ·
  1

• feat(generative_reward_model): add GenRM training resources server

  area:environmentcomplexity:mediumfeatureneeds-review
  #3635·HeyyyyyyG opened Sep 22, 2026·

  ·

• feat(checkpoint): resume proof refinement agent turns

  area:agentcomplexity:mediumfeatureneeds-review
  #3634·zyzhou5 opened Sep 22, 2026·

  ·
  1

• feat(core): expose aiohttp pool wait telemetry

  area:corefeature
  #3632·athitten opened Sep 22, 2026·

  ·
  1

  3

  

• [Do not merge] Fsiino/inkling tb21 accuracy gap

  #3629·fsiino-nvidia opened Sep 22, 2026·

  ·
  1