fix(tb): E+C empty-gold scoring + unknown abstention finalize - #2833
Draft
Dominic Nguyen (datduyng) wants to merge 6 commits into
Draft
fix(tb): E+C empty-gold scoring + unknown abstention finalize#2833Dominic Nguyen (datduyng) wants to merge 6 commits into
Dominic Nguyen (datduyng) wants to merge 6 commits into
Conversation
- Dispatcher: return unknown on DispatcherName instead of throwing in finalizeAction / canTranslateWithoutContext (preserves multi-action). - Runner scoring: unknown schema-match → zero-action PASS on empty gold (E); filter chat.generateResponse / utility.claudeTask from scored chosen (C). - Share HARDCODED_NON_EVAL_ACTION_IDS from eligibleActions with runner. - Unit tests: 10 cases covering abstention, non-eval, real errors still fail. Live claim (separate run): unknown-throw 1201→0, neg_pass 0.13→0.618 on 5832 cells.
- Catch-path abstention rawChosen uses schemaName dispatcher (matches finalize). - Package export chain index → runner → scoring (drop dual identical barrels).
Dominic Nguyen (datduyng)
marked this pull request as draft
August 9, 2026 06:36
- approve-and-eval.mjs drives the dual-root eval and, on completion, publishes eval-report.html + eval-results.json and regenerates the landing repo (index + per-cell explorer) - config via commander flags / --config <file.json> instead of TB_* env - no hardcoded paths; runtime-ts defaults to the local worktree
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
unknownonDispatcherNameinstead of throwing; TB scoring treats unknown schema-match as zero-action PASS on empty-gold (not harness JSON error).HARDCODED_NON_EVAL_ACTION_IDS(chat.generateResponse,utility.claudeTask); runner drops them from scored chosen while keeping raw for viz.Live proof (separate claim run)
5832 cells · ~$185 · art
ec-live-gen1k· goal note [[20260808-205309-experiment-typeagent]]Test plan
jest --testPathPattern=runnerScoring→ 10/10@typeagent/benchmarks+agent-dispatcher