Add route_error state-metrics scenario (model-error routing) - #7
Merged
Conversation
Scenario #3 per David's 'modify the BPS model, not the log' instruction: generate GT with the correct case_type routing, then continue the ongoing cases with a model whose XOR routing is inverted, and detect the deviation. - tools/generate_route_error.py: case_type-driven XOR chain with structurally-distinct but timing-symmetric branches (correct routing baked in via branch_rules). - perturb.build_route_error_params: invert red/blue routing on the first N gateways (level 0 = verbatim copy). - Wired as a simulator-driven param family (dispatch in _prepare_params_for_level), DatasetSpec synthetic_route_error, CLI choice. - Unit tests for the builder. - Docs (PAPER_SCENARIOS.md / EXPERIMENT_SCENARIOS.md) synced to David's final scenario set: drop old XOR-redistribution / temporal-drift, redesign #4 to compensating front/back branches. Validated (Scope A, 5 runs, invert 0-3 gateways): timing baselines (cycle_time/red/rtd) exactly blind; state activity_case top metric (0.933); ngd_n2 co-detects (0.917) but ranks below and is blind at low levels.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Scenario #3 — model-error routing
Implements David's modify the BPS model, not the log instruction. Generate ground truth with the correct
case_typeXOR routing, then continue the ongoing cases with a model whose routing is inverted, and ask whether the metric detects the deviation.What's here
tools/generate_route_error.py—case_type-driven XOR chain (default 3 gateways) with structurally-distinct but timing-symmetric branches; correct routing (red→A, blue→B) baked in viabranch_rules.perturb.build_route_error_params— inverts red/blue routing on the first N gateways; level 0 = verbatim copy._prepare_params_for_level),DatasetSpec synthetic_route_error, CLI--perturbation route_error.Validation (Scope A, 5 runs, invert 0–3 gateways)
activity_case(pairing)activity_typengd_n2cycle_time/red/rtdThe core SOTA timing-aggregation metrics are dead blind (timing-symmetric branches); the per-case pairing metric
activity_caseis the single top-ranked metric. Basic sim confirmed zero-crossover routing.Caveat:
ngd_n2partially co-detects because real re-routing genuinely shifts activity composition — perfect ngd-blindness stays with the pure tag-swap oracles (case_route/label_swap).