Skip to content

TS-Reasoner v19.0.0: TS-AGL Tiny Learned Router

Choose a tag to compare

View all tags
@BoggersTheFish BoggersTheFish released this 01 Jun 00:03
· 14 commits to main since this release
v19.0.0
9cab782
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

TS-Reasoner v19.0.0 adds the TS-AGL Tiny Learned Router.

Release claim:

  • TS-AGL can train a tiny dependency-free router from the v18 trace-mined dataset.
  • The router predicts operation routes or abstains.
  • Heldout routing and abstention are evaluated.
  • Routed calls still go through TSCall/risk/adapters/verifier boundaries.
  • Learned confidence is not proof.
  • Candidate graph contamination remains zero.
  • No external LLM is used.

Boundary:

  • learned routing only
  • no proof authority
  • no external LLM
  • no neural dependency
  • confidence is not proof
  • routed calls still use typed TSCall/risk/adapters/verifier boundaries
  • candidate graph contamination remains zero

Verification:

  • python3 -m unittest discover -q
  • python3 scripts/validate_ts_agl_domain_packs.py
  • python3 scripts/build_ts_agl_router_dataset.py
  • python3 scripts/evaluate_ts_agl_router_dataset.py
  • python3 scripts/train_ts_agl_tiny_router.py
  • python3 scripts/evaluate_ts_agl_tiny_router.py
Assets 25
Loading