All pull requests

Pull requests

• fix(ci): size context work budget for bounded embeddings

  #167·asutermo opened Oct 30, 2026Collaborator·

  ·

• feat: screen verifier cheats and measure native controls in CI

  #164·asutermo opened Oct 30, 2026Collaborator·

  ·
  26

• test(metric-benchmark): revalidate sealed jobs after validation failure

  #163·mstaats-nvidia opened Oct 30, 2026Contributor·

  ·

• feat(metric-benchmark): prototype Helix submission and evidence collection

  #162·mstaats-nvidia opened Oct 30, 2026Contributor·

  ·
  1

• feat(generation-benchmark): run generation experiments on metric coverage

  #161·bcleary-nv opened Oct 29, 2026Contributor·2/2·7 tasks·

  ·
  7

• docs(generation-benchmark): describe the generation experiment

  #160·bcleary-nv opened Oct 29, 2026Contributor·1/2·7 tasks·

  ·

• docs(metric-benchmark): sketch Helix submission and result collection

  #159·mstaats-nvidia opened Oct 29, 2026Contributor·

  ·
  1

• feat(first-eval): offer local trajectory saving before run settings

  #153·jessbuhl opened Oct 28, 2026Collaborator·

  ·
  22

• docs(metric-benchmark): locate results validation in the stage list

  #151·mstaats-nvidia opened Oct 28, 2026Contributor·

  ·

• docs(audit): separate tool metadata from Ethos requirements

  #114·jessbuhl opened Oct 22, 2026Collaborator·

  ·
  2

• feat(trace-environment)!: read Compass Trace instead of ATIF

  #109·aleckhoury opened Oct 21, 2026Collaborator·

  ·
  3

• refactor!: load traces only through NeMo Compass; remove ATIF converters

  #108·aleckhoury opened Oct 21, 2026Collaborator·

  ·

• feat(trace-environment): record tool calls across many traces in a corpus

  #100·aleckhoury opened Oct 20, 2026Collaborator·

  ·
  1

• feat(metric-benchmark): orchestrate configured experiments through stage libraries

  #99·mstaats-nvidia opened Oct 20, 2026Contributor·5/5·

  ·
  12

• fix(ethos): remove the required tool inventory

  #69·jessbuhl opened Oct 13, 2026Collaborator·

  ·
  4

• feat(eval-author): add a smoke-task template for environment kits

  #53·aleckhoury opened Oct 2, 2026Collaborator·5 of 6 tasks·

  ·

• docs(eval-author): allow post-audit continuation beyond tool gaps

  #10·aleckhoury opened Sep 16, 2026Collaborator·

  ·