feat(bfcl): add BFCL v4 edge-agentic accuracy + performance integration#346

Open

Palanivelg wants to merge 30 commits into

mlcommons:mainfrom

Palanivelg:feat/bfcl-v4-combined

Commits on Jun 3, 2026

Commits on Jun 8, 2026

feat(bfcl): add BFCL v4 single-turn + multi-turn accuracy integration
Palanivelg
committed
feat(bfcl): per-category sampling + tiny-subset floor + Thor example
Palanivelg
committed
docs(bfcl): generalize "Thor" to "edge device" in BFCL v4 example
Palanivelg
committed
fix(bfcl): guard rt_settings access in accuracy-only mode
Palanivelg
committed
fix(bfcl): per-phase drain timeout; accuracy phases drain unbounded
Palanivelg
committed
fix(bfcl): stop duplicating tool_calls in TextModelOutput.output
Palanivelg
committed
fix(bfcl): score against structured tool_calls to tolerate prose preambles
Palanivelg
committed
fix(bfcl): send tool_choice="auto" for single-turn function calling
Palanivelg
committed
feat(bfcl): forward seed parameter through adapters and MT runner
Palanivelg
committed
docs(bfcl): add reproduction guide with Thor validation results
Palanivelg
committed
docs(bfcl): rewrite README as a beginner-friendly step-by-step guide
Palanivelg
committed
docs(bfcl): update README with corrected Thor validation data
Palanivelg
committed
fix(bfcl): add soundfile dependency for qwen_agent transitive import
Palanivelg
committed
refactor(bfcl): rename example folder to 10_Edge_Agentic_Example; add run_accuracy.sh
Palanivelg
committed
fix(bfcl): address code review issues in execution, runner, and scorer
Palanivelg
committed
docs(bfcl): annotate YAML with do-not-change submission rules
Palanivelg
committed
fix(bfcl): resolve CI failures — numpy pin, templates, mypy, audit
Palanivelg
committed
fix(bfcl): honor configured accuracy_timeout_s for accuracy phase drain
Palanivelg
committed
fix(bfcl): run_accuracy.sh — render temp config instead of invalid flags
Palanivelg
committed
Merge branch 'pr-331' into feat/edge-agentic-perf
Palanivelg
committed

Commits on Jun 11, 2026

feat(examples): add edge agentic-coding performance benchmark with inline online checker
Palanivelg
committed

Commits on Jun 15, 2026

feat(examples): add edge agentic-coding 1h/2.5h perf datasets + configs
Palanivelg
committed

Commits on Jun 22, 2026

Commits on Jun 23, 2026