Skip to content

Add BFCL cohesive namespace candidate ablation - #42

Closed
SonAIengine wants to merge 1 commit into
codex/bfcl-selection-guidance-ablationfrom
codex/bfcl-cohesive-namespace-ablation
Closed

Add BFCL cohesive namespace candidate ablation#42
SonAIengine wants to merge 1 commit into
codex/bfcl-selection-guidance-ablationfrom
codex/bfcl-cohesive-namespace-ablation

Conversation

@SonAIengine

Copy link
Copy Markdown
Owner

Summary

  • add --cohesive-namespace-candidates to BFCL model loop and sweep runners
  • keep raw retrieval unchanged while compressing the LLM-facing tools_presented set for multi-action retrieved queries with a cohesive dotted namespace family
  • document the first qwen3.6-27B small smoke where the xgen-0.27 milestone gate passes

qwen3.6-27B Smoke Evidence

Via go165 vLLM qwen3.6-27b, category limit 5, row/retrieved k=5.

Artifact: /tmp/gtc-bfcl-qwen027-cohesive-smoke-v2.json

  • row exact: 1.00
  • retrieved exact: 0.95
  • retrieval recall: 1.00
  • row-source preservation: 0.95
  • parallel_multiple exact: 0.80
  • milestone gate: pass

Hard-case bundle: /tmp/gtc-bfcl-qwen027-cohesive-hardcases-v2

  • cases: 1
  • remaining failure: parallel_multiple_4 integral sibling ambiguity

This is small-smoke evidence, not a publishable full BFCL number. Next promotion step is a failure-subset run and then a larger sweep.

Validation

  • poetry run ruff check .
  • poetry run ruff format --check .
  • poetry run pytest tests/ -q -> 952 passed, 3 skipped

@SonAIengine

Copy link
Copy Markdown
Owner Author

Closing this internal stacked PR as superseded by the current main branch. The branch is preserved for history.

@SonAIengine SonAIengine closed this Aug 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant