bill/whetstone_DSL
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add adaptive raw top-gap retry loop (sprint 259)

Browse Source
This commit is contained in:
Bill
2026-02-26 16:33:19 -07:00
parent df0de8156b
commit ba1e0dc3e8
7 changed files with 141 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docs/constructive_editing_runtime_plan.md
View File

@@ -228,3 +228,7 @@ Planning runtime controls (active):
- `WSTONE_NATIVE_RAW_TOP_GAP_REQUIREMENTS`
- `WSTONE_NATIVE_RAW_TOP_GAP_MAX_SIGNALS`
- `native_raw_top_gap_requirements`
- raw candidate search can run adaptive retry with expanded top-gap requirements:
- `WSTONE_NATIVE_RAW_TOP_GAP_ADAPTIVE_RETRY`
- `WSTONE_NATIVE_RAW_TOP_GAP_RETRY_SIGNALS`
- `native_raw_adaptive_retry`

2
docs/generator_readiness_gap_registry_2026-02-26.md
View File

@@ -47,7 +47,7 @@ This is the canonical dated registry for "not production-ready" generator gaps.
| GR-018 | Performance/security/rollout constrained refactor enforcement gap | `docs/gap_hunt_fullstack_multifile_2026-02-26.md`, `logs/taskitem_runs/challenging_fullstack_multifile_20260226_r5/results.jsonl` | `partial` | Hard checks now enforce migration rollback + data-loss policy, security deny-by-default, SLO p95 presence, and rollout staged+abort policy. Enforcement is contract-level; generator capability under these constraints is still weak in C++ AB path. | Constraint-aware generation follow-up |
| GR-019 | Parity-blocked readiness load (gating without capability closure) | `logs/taskitem_runs/challenging_fullstack_multifile_20260226_r7/summary.json`, `logs/taskitem_runs/challenging_subset_prod_20260226_r7/summary.json`, `docs/sprint225_227_execution_tracker_2026-02-26.md` | `partial` | Sprint 225-227 reduced blocked parity load from `6` to `0` on both tracked hard catalogs while keeping unresolved divergence at `0`. This closes immediate safety debt for current corpora, but robustness is still contingent on pattern-driven repair classes. | Generalize repairs beyond queue-shaped transpile outputs |
| GR-020 | Semantic fallback overuse masks weak native decomposition | `logs/taskitem_runs/TEST_ONLY_sprint236_semantic_fallback_audit_20260226/semantic_fallback_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint237_semantic_fallback_gate_fail_20260226/semantic_fallback_budget_gate.json`, `logs/taskitem_runs/TEST_ONLY_sprint238_native_gate_fail_20260226.json`, `logs/taskitem_runs/TEST_ONLY_sprint239_semantic_fallback_audit_20260226/semantic_fallback_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint240_retry_20260226_145631/00_summary.json`, `docs/sprint236_execution_tracker_2026-02-26.md`, `docs/sprint237_execution_tracker_2026-02-26.md`, `docs/sprint238_execution_tracker_2026-02-26.md`, `docs/sprint239_execution_tracker_2026-02-26.md`, `docs/sprint240_execution_tracker_2026-02-26.md` | `partial` | Sprint 236 added deterministic fallback-gap auditing with tool + constraint metadata. Sprint 237 added fallback-budget hard gate (`max-fallback-rate`) and produced expected fail/pass artifacts. Sprint 238 added native decomposition hard gate in pipeline (`min task count`, `min semantic signal count`) so weak native generation can be blocked before fallback masking. Sprint 239 added native reason enrichment and improved semantic signal density (`0 -> 6`). Sprint 240 added native decomposition retry with explicit minimum-task policy; on current sample retry attempted but did not improve depth (`2 -> 2`). | Native decomposition task-depth upgrade (increase native task granularity beyond 2) |
| GR-021 | Impact-specific native decomposition coverage not enforced uniformly | `docs/native_decomposition_impact_list_2026-02-26.md`, `tools/mcp/profiles/native_decomposition_impact_profiles.json`, `logs/taskitem_runs/TEST_ONLY_sprint241_fullstack_impact_20260226_150416/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint241_native_impact_aggregate_20260226/native_impact_coverage_aggregate.json`, `logs/taskitem_runs/TEST_ONLY_sprint242_impact_remediation_loop_20260226/remediation_loop_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint243_impact_remediation_tasks_loop_20260226_r2/remediation_loop_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint244_autofill_20260226_151651/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint244_autofill_enforce_20260226_151709/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint245_intrinsic_20260226_151835/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint246_multishot_20260226_153033/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint246_multishot_enforce_20260226_153045/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint247_singleshot_20260226_153230/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint247_singleshot_enforce_20260226_153241/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint248_rawsearch_20260226_153903/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint249_rawguard_20260226_154028/02ae_raw_candidate_search.json`, `logs/taskitem_runs/TEST_ONLY_sprint250_closure_ladder_20260226/closure_ladder_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint251_rawvariants_20260226_154355/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint252_closure_ladder_policy_skip2_20260226/closure_ladder_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint253_closure_ladder_batch_20260226/closure_ladder_batch_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint254_closure_ladder_batch_20260226_r2/raw_gap_backlog.json`, `logs/taskitem_runs/TEST_ONLY_sprint255_rawharden_20260226_161924/00_summary.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_162423/00_summary.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_162633/02ae_raw_candidate_search.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_162924/01e_raw_top_gap_requirements_report.json` | `partial` | Sprint 241 introduced impact profiles/gates; 242-243 remediation loops; 244 first-pass autofill; 245 intrinsic constraints only (no uplift); 246 multishot closure; 247 single-shot shaping closure; 248 raw candidate search no uplift; 249 no-uplift guardrail; 250 closure ladder; 251 richer raw variants no uplift; 252 history-aware ladder routing; 253 batch ladder analytics; 254 raw gap backlog synthesis from `raw_only` attempts identifies top missing raw signals; 255 adds first-pass raw top-gap hardening that closes hard-sample profile coverage (`failing_profile_count 7 -> 0`) by injecting missing ops/contracts/reasons and minimum task depth; 256 adds top-gap weighted candidate selection telemetry and tie-break policy; 257 adds hard fail policy when top-gap uplift is required but absent (`exit 18`); 258 injects backlog-driven top-gap normalized requirements before raw generation, but measured intrinsic output still shows no uplift on this sample (`best_top_gap_score=0`, `failing_profile_count=1`). | Add adaptive raw retry loop that iteratively re-synthesizes and expands top-gap requirements when uplift remains zero |
| GR-021 | Impact-specific native decomposition coverage not enforced uniformly | `docs/native_decomposition_impact_list_2026-02-26.md`, `tools/mcp/profiles/native_decomposition_impact_profiles.json`, `logs/taskitem_runs/TEST_ONLY_sprint241_fullstack_impact_20260226_150416/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint241_native_impact_aggregate_20260226/native_impact_coverage_aggregate.json`, `logs/taskitem_runs/TEST_ONLY_sprint242_impact_remediation_loop_20260226/remediation_loop_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint243_impact_remediation_tasks_loop_20260226_r2/remediation_loop_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint244_autofill_20260226_151651/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint244_autofill_enforce_20260226_151709/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint245_intrinsic_20260226_151835/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint246_multishot_20260226_153033/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint246_multishot_enforce_20260226_153045/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint247_singleshot_20260226_153230/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint247_singleshot_enforce_20260226_153241/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint248_rawsearch_20260226_153903/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint249_rawguard_20260226_154028/02ae_raw_candidate_search.json`, `logs/taskitem_runs/TEST_ONLY_sprint250_closure_ladder_20260226/closure_ladder_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint251_rawvariants_20260226_154355/00_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint252_closure_ladder_policy_skip2_20260226/closure_ladder_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint253_closure_ladder_batch_20260226/closure_ladder_batch_summary.json`, `logs/taskitem_runs/TEST_ONLY_sprint254_closure_ladder_batch_20260226_r2/raw_gap_backlog.json`, `logs/taskitem_runs/TEST_ONLY_sprint255_rawharden_20260226_161924/00_summary.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_162423/00_summary.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_162633/02ae_raw_candidate_search.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_162924/01e_raw_top_gap_requirements_report.json`, `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/02af_raw_adaptive_retry_score.json` | `partial` | Sprint 241 introduced impact profiles/gates; 242-243 remediation loops; 244 first-pass autofill; 245 intrinsic constraints only (no uplift); 246 multishot closure; 247 single-shot shaping closure; 248 raw candidate search no uplift; 249 no-uplift guardrail; 250 closure ladder; 251 richer raw variants no uplift; 252 history-aware ladder routing; 253 batch ladder analytics; 254 raw gap backlog synthesis from `raw_only` attempts identifies top missing raw signals; 255 adds first-pass raw top-gap hardening that closes hard-sample profile coverage (`failing_profile_count 7 -> 0`) by injecting missing ops/contracts/reasons and minimum task depth; 256 adds top-gap weighted candidate selection telemetry and tie-break policy; 257 adds hard fail policy when top-gap uplift is required but absent (`exit 18`); 258 injects backlog-driven top-gap normalized requirements before raw generation; 259 adds adaptive retry with expanded top-gap requirements and deterministic retry scoring. Measured intrinsic output still shows no uplift on this sample (`retry_top_gap_score=0`, `retry_fail=1`, `applied=false`). | Add model/prompt-level intrinsic generation intervention keyed by missing-signal classes (beyond requirement text injection alone) |
## What Was Covered Today (Sprints 175-184)

16
docs/progress_log_2026-02-26.md
View File

@@ -458,3 +458,19 @@
- Current measured result on hard sample:
- injection selected `5` top signals from `20`
- intrinsic raw outcome unchanged (`selected_variant=0`, `best_top_gap_score=0`, `failing_profile_count=1`).
## Sprint 259 Added (Same Day)
- Added adaptive raw top-gap retry loop:
- `WSTONE_NATIVE_RAW_TOP_GAP_ADAPTIVE_RETRY`
- `WSTONE_NATIVE_RAW_TOP_GAP_RETRY_SIGNALS`
- summary packet:
- `native_raw_adaptive_retry`
- Retry executes when weighted top-gap score has no uplift, then applies only if better.
- Dated A/B artifacts:
- OFF: `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163143/00_summary.json`
- ON: `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/00_summary.json`
- retry detail: `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/02af_raw_adaptive_retry_score.json`
- Current measured result on hard sample:
- retry attempted and succeeded operationally
- no intrinsic uplift (`retry_top_gap_score=0`, `retry_fail=1`, `applied=false`).

43
docs/sprint259_execution_tracker_2026-02-26.md Normal file
View File

@@ -0,0 +1,43 @@
# Sprint 259 Execution Tracker - 2026-02-26
## Scope
- `sprint259_plan.md`
## Implemented
Pipeline adaptive retry in `tools/mcp/run_sprint_taskitem_pipeline.sh`:
- Added:
- `WSTONE_NATIVE_RAW_TOP_GAP_ADAPTIVE_RETRY`
- `WSTONE_NATIVE_RAW_TOP_GAP_RETRY_SIGNALS`
- Retry trigger condition:
- weighted top-gap mode enabled
- `best_top_gap_score <= baseline_top_gap_score`
- Retry behavior:
- synthesize expanded top-gap requirements
- rerun raw task generation
- score retry candidate
- apply retry output only if it improves fail count, top-gap score, or task count tie-break
- Added summary packet:
- `native_raw_adaptive_retry`
## Validation Artifacts
Adaptive retry OFF:
- `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163143/00_summary.json`
Adaptive retry ON:
- `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/00_summary.json`
- retry artifacts:
- `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/02af_raw_adaptive_retry.json`
- `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/02af_raw_adaptive_retry_score.json`
- `logs/taskitem_runs/01a_fallback_intake_spec_20260226_163148/02af_raw_adaptive_retry_requirements_report.json`
Observed result on hard sample:
- retry attempted successfully
- `retry_top_gap_score=0`
- `retry_fail=1`
- `applied=false` (no uplift to apply)
## Explicit Completion Signal
- Sprint 259: `DONE` (implemented + validated; no uplift on this sample)