GitDuck Bench V1.1.0 — The Geese Have Rankings

V1.1.0 adds reproducible derived views over the immutable V1 result corpus.
No benchmark case, scorer contract, or historical attempt record was changed.

28 attempts · 0 ducks
Distance-to-Duck leaderboard
Outcome-awareness diagnostic
6 exact-text collision events across 5 shared digests

Finding the load-bearing key is not equivalent to providing exact
provenance. The awareness table preserves detection; the leaderboard ranks
precision.

Distance to Duck          = unofficial derived leaderboard
Outcome Awareness         = diagnostic projection, not a ranking
DUCK 1/1                  = only official benchmark metric
🔑                        = badge, never a sorting bonus
Δexact                    = defined only for eligible correct-action answers

Distance-to-Duck leaderboard
— Δexact is the multiset symmetric difference over all exact-set fields
of the answer contract (support keys + evidence refs + blocked set), defined
only when schema, authorization and action selection all PASS, so Δexact = 0
⇔ DUCK 1/1 — asserted per row by the verifier. It reads as an edit distance:
the number of single-element fixes to the exact support contract. Dense
ranks; chronology never breaks ties; byte-identical attempts share a rank
but stay separate attempts; raw-model and agent-product lanes never mix.

Closest attempt       Claude Sonnet 5 — Δ 1 (one extra related support key)
Best clean miss       nemotron-3-nano:30b — Δ 2 (one wrong key, zero fabricated refs)
Outcome-key citers    4/28; first: minimax-m2.7 (attempt 12)

The Outcome-awareness projection (second table of the leaderboard) keeps
detection visible without rewarding it: four attempts really did see the
load-bearing key; Sonnet used it almost correctly, two key-bearers caught it
in a net along with the whole pond.

Exact-byte convergence
— the collision ledger ranked as events, never as synthetic pairs:
6 collision events · 5 shared answer digests · 4 cross-provider · 1
intraprovider · 1 original same-infrastructure event that remains UNVERIFIED
FOREVER. Exact bytes do not prove copying, distillation, caching, aliasing,
or a shared physical backend.

Both tables are generated from the published records and regenerate
byte-identically:

node tools/verify-manifest.mjs
node tools/verify-results.mjs
node tools/verify-derived-tables.mjs

Manifest SHA-256:

9db6563b55d3d4c6dcb6316306a72073e23cc910d4669471a2091992c66a5278

Also in this release: the goose is drawn as a swan 🦢 — the goose glyph
U+1FABF is Unicode 15 and renders as a box on Windows 10 and other frozen
emoji fonts. The canonical award name is unchanged: GOOSE everywhere data
speaks; the swan is its well-supported stage mask.

GitDuck Bench V1.1.0 — The Geese Have Rankings. 🦆🦢📊

GitDuck Bench V1 — The Pond Is Empty

One frozen case. 28 attempts. 0 ducks. 6 exact-text collision events.

GitDuck Bench measures one competence: exact action selection AND exact
support provenance on a real Git decision, scored by a deterministic offline
scorer — no judge model, no rubric, no partial credit. The official metric is
binary: 🦆 DUCK = all four axes PASS.

V1 board: 27 raw-model attempts (18 Ollama-cloud open models; raw APIs from
OpenAI, Anthropic ×4, xAI, Google ×3) plus 1 agent-product attempt. Nearly
every attempt picked the correct action; none proved it with the exact
evidence set. Closest: Claude Sonnet 5, one extra support key from 1/1. Six
byte-identical wrong answers across independent ecosystems — case-induced
exact convergence, sometimes byte for byte.

One case is not a global model ranking. Boundaries:
reports/v1/limitations.md.

Verify everything offline — no keys, no network:

node tools/verify-manifest.mjs
node tools/verify-results.mjs

results/v1/manifest.json pins the SHA-256 of every published file and
excludes itself from its own inventory. Manifest SHA-256:

9e853841f8a5b98a3f8537d5948916fc493c2c1a7100dee0c1b06cecd2c4649b

Licenses: Apache-2.0 (code) · CC BY 4.0 (case, data, reports). If your
attempt scores 1/1 under the frozen contract, open an issue titled
DUCK CLAIM — the scorer decides, not us.

GitDuck Bench V1 — The Pond Is Empty.
Different ecosystems. Same way to goose. Sometimes byte for byte. 🦆🦢