Stigmergy ablation
Method: In-tree honest proxies + upstream meeting-sched fixture. LLM repro optional via OLLAMA_URL.
Summary
| Bench | Stigmergy / pressure-field | Random / conversation | Notes |
|---|---|---|---|
Chain dispatch (stigmergy_ablation) | 5/5 zones | 0/5 zones | Dependency wake via pheromone deposits |
| Gridworld success rate (8 runs) | 100% | 100% | Both configs reach target; chain dispatch is the discriminant |
Meeting-sched proxy (govcraft_acceptance) | 41.7% booking rate | 8.3% booking rate | Discrete slot scheduling with PheromoneField |
| Upstream fixture | deterministic PF / conv | — | fixtures/upstream_easy_seed4242.json |
| Published LLM reference | 48.5% (reference) | 12.6% (reference) | Honest gap unless OLLAMA_URL repro lands in band |
Honest gap table (Bet 1)
| Metric | Measured (in-tree) | Published reference |
|---|---|---|
| Upstream task shape pressure-field (deterministic) | ~80% solve (5 trials, seed 4242–4246) | 48.5% LLM |
| Upstream task shape conversation (deterministic) | ~100% solve (same trials) | 12.6% LLM |
| Slot proxy stigmergy (decay on) | ~41.7% booking rate | — |
| Slot proxy stigmergy (decay off) | ~8.3% booking rate | — |
| Slot proxy random | 8.3% booking rate | — |
Decay ablation: Sequential within-round booking used to mask decay (always filled all 5 slots). Fixed with parallel picks + conflict resolution: decay off collapses to one slot (~8%); decay on spreads (~42%). cargo run -q -p govcraft_acceptance -- --ablation validates ≥5% gap.
Upstream fixture
cargo run -q -p govcraft_acceptance -- --mode upstream_meeting_sched
cargo run -q -p govcraft_acceptance -- --upstream-repro
Fixture: ScheduleGeneratorConfig::easy(), seed 4242, at benches/govcraft_acceptance/fixtures/upstream_easy_seed4242.json.
Honest labeling
- In-tree proxy:
benches/govcraft_acceptanceandbenches/stigmergy_ablationuse CastellanPheromoneField+ scheduler semantics. - Upstream fixture: rooms, attendees, duration slots, overlap predicate (distinct from slot proxy).
- Not reproduced (LLM): 48.5% vs 12.6% unless
--upstream-reprowithOLLAMA_URLmeasures in-band rates. - CI: Slot proxy + ablation in
./scripts/verify.sh; upstream mode opt-in.
Provenance
Upstream meeting-sched task shape and published LLM band cite pressure-field-experiment. Crate/binary names (govcraft_acceptance, GOVCRAFT_*) are historical identifiers.
Reproduce
cargo run -q -p stigmergy_ablation
cargo run -q -p govcraft_acceptance
cargo run -q -p govcraft_acceptance -- --mode upstream_meeting_sched
bash docs/demo/run-ablation.sh
./scripts/verify.sh