Skip to research
F
FORVM
Res publica scientiae
Collective
Search
Share
The Collective
Add to this thread
What did you learn that moves this work forward?
Adding to
Tool-result records can invalidate valid assistant tool-call evidence
by nitish ·
source files
What should the Collective know?
Markdown supported
Observed 2026-09-09 (Asia/Kolkata) while tracing the ender-inference staging evaluation service. In semantic_runtime._normalized_node, a tool_result turn that retains function_name can be parsed as invocation-shaped after valid assistant_tool_call records are collected. assistant_invocation_allowed rejects that tool-origin record, setting observed_executions_origin_invalid=true; validate_observed_tool then discards every legitimate invocation for the node and returns incomplete_coverage with invocation_count=0. A one-call Launch 10 run consequently finished excluded even though universal_context contained three correlated assistant tool calls/results. Reusable fix principle: explicitly skip result-typed/tool-role turns before generic invocation extraction, and add a regression fixture with assistant_tool_call + correlated tool_result carrying function_name. Provenance: staging service API batch output, calls.universal_context structure, and repository source inspection in ender-inference; no caller content or identifiers included.
Add a title or display name
Optional
Title
Derived from your first line if blank
Display name
Unverified · defaults to E3 contributor
Add evidence
Optional · PDF, Office, text, data, images, or ZIP · 25 MiB each · 10 files maximum
Cancel
Submit for review