$0.10/1k-q overstated the qwen-27B case. Honest range: ~$0.07-0.16 per
1,000 queries of GPU electricity — hermes-8B $0.07-0.09 (under a dime),
qwen-27B $0.12-0.16 (over a dime; claim_lattice dearer than quote from
more prefilled context). Fixes the §5.4 'either rig' claim.
The report is now cost AND value (quality-per-dollar), not cost-only.
§5.5 quality delta: substrate-vs-solo (code judge, n=30) on numeral +
stale fixtures, hermes-8B vs qwen-27B-nothink. Substrate lifts both
2-13x and they nearly converge (hermes 18/21, qwen 19/26 grounded-
correct) — grounding comes from retrieval+verification, not parametric
size, so the base model matters far less. Bare-model failure styles
diverge (hermes abstains, qwen fabricates). Quality-per-dollar: ~18-26/30
grounded for $0.085-0.158/1k-q.
Honest caveats recorded: CORRECT=grounded-in-2010-corpus (not current);
the stale fixture is no longer post-corpus; the qwen thinking bug (39c040c)
that voided the first run and the judge strengthening (2d31866). §1
reframed; §8 repro + §9 next updated (post-2010 fixture, SOTA judge for
residue).
The code judge bailed to JUDGE_ERROR on 40% of in-corpus answers: HYBRID
(partial grounding) with low NLI entail, where the entity-grounding
rescue needs ZERO unsourced specifics. A single extra proper noun
('Emperor Honorius', 'Alexander Molossus' — an alias/paraphrase) blocked
rescue even with verbatim quotes verified and the answer correct.
New HYBRID resolution tier: rescue to CORRECT_GROUNDED when the verifier
confirmed >=1 verbatim quote, the subject anchor is in gold (on-topic),
there is NO unsourced NUMERIC specific (wrong dates/counts stay residue),
and NLI isn't strongly contradicting. Unsourced proper nouns are treated
as aliases/paraphrase; unsourced numerics (the real factual-error class)
keep the answer as JUDGE_ERROR. Validated on the 12 real residue cases:
9 -> CORRECT (all genuinely right), 3 stay residue (unsourced numerics).
JUDGE_ERROR 40% -> ~10%. self-test 4/4; 2 new tier tests; suite 2549.
Root cause of 'arborist abstains on everything with qwen' (fox 2026-05-21):
Qwen3 thinking-on default burns the entire token budget on hidden <think>
reasoning over a 20K RAG context and returns EMPTY message.content
(measured: 768/768 completion tokens, content '') -> every arborist answer
UNGROUNDED. Bench harnesses passed enable_thinking=False via the MODELS
dict, but the CLI + control_ab did not, so the quality bench was measuring
a thinking-budget-exhaustion artifact, not abstention.
OpenAICompatibleClient now defaults Qwen3 to enable_thinking=False unless a
caller set it explicitly (reasoning-variant path passes True, preserved).
Verified: same France query goes empty/UNGROUNDED -> STRICT 'Nicolas
Sarkozy' with the flag. Fixes every caller (CLI, control_ab). 5 tests;
full suite 2547 passed. Today's qwen QUALITY numbers are void and need
re-running; energy numbers stand (real inference happened regardless).
Was a stale hardcoded 'same Hermes; judge=Opus hermetic' label that
misreported any run with --model/--judge overrides (e.g. qwen + code
judge). Now reflects the real config — honest header for the artifact.
fox 2026-05-21: two corpora exist (2003-05-16 + 2010); the 2010 one is
loaded in ~/.arborist/shards (verified: 866K docs/shard, has Barack
Obama / YouTube articles a 2003 dump can't). CLAUDE.md Live-endpoints and
the energy-COGS report now state 2010; fabrication-bait fixtures target
post-2010 events. Historical closed-ticket prose left as point-in-time
records.
- §2 now a two-rig table: 4090/Qwen-27B-Q4_K_XL/llama.cpp (isolated) vs
3090/Hermes-8B-FP8-Dynamic/vLLM (live/public). Spells out that the
cross-model rate gap confounds FOUR variables (params, quant/precision,
engine, GPU) — not '8B vs 27B' alone.
- §5.4 hermes results under live traffic. Method finding: the watt_bench
window integral is unusable on a contended card, but the slope
calibration survives clean (cancels the variable shared-load baseline)
— so on shared/public cards use watt_calibrate, not the window integral.
Rates: hermes 0.109/4.40 J/tok vs qwen 0.175/6.16; decode 35-40x prefill.
- Per-query DOLLARS both rigs @$0.33/kWh: a grounded substrate answer is
<$0.10 per 1000 queries (hermes $0.085, qwen $0.158); hermes ~half qwen.
- §5.2 budget confound corrected to flag Hermes-tuning honestly (was
rationalized). §9: fixed-budget apples-to-apples re-run + per-model
budget tuning added as next steps.
Full report for Dav1d: qwen-nothink on the dedicated 4090. Headline —
prefill 0.175 J/input-tok ($0.016/M) vs decode 6.157 J/output-tok
($0.564/M), decode 35x dearer per token. The substrate prefills ~6.6k
input tok/query (vs solo ~52, 127x): ~67% of its GPU energy is reading
the retrieved context, not generating. Per-query 1719 J (substrate-CL)
vs 95 J (solo) = $0.16 vs $0.009 per 1k queries; calibration predicts
measured within ~5%. Real cost is CPU latency, not GPU watts.
Documents the full methodology + the corrections that got here
(contamination -> isolation; blended mean-W -> measured states; gross
-> not-attributable; per-token -> separated input/output; len//4 ->
real usage), threats to validity, reproducibility (commits, config hash,
persisted samples), and next rigs (3090+Hermes, reasoning, prefill-cache
study).
fox 2026-05-21: account wattage for input and output separately. Prefill
(process all prompt tokens, parallel/compute-bound) and decode (generate
output, autoregressive/bandwidth-bound) are different GPU ops with
different J/token — a single per-token number can't represent both.
Slope calibration (no sub-request power alignment): sweep prompt length
at tiny max_tokens -> prefill J/input-tok (fixed overhead cancels in the
slope); fix a tiny prompt and sweep forced output length (ignore_eos) ->
decode J/output-tok. Prefill kept COLD (unique filler so cached_tokens=0).
Reuses watt_bench probes. Bad points (context overflow) skip, not abort.
Measured qwen-nothink/4090 @$0.33/kWh: prefill 0.175 J/tok
($0.016/M-input-tok), decode 6.16 J/tok ($0.564/M-output-tok) — decode
35x dearer per token. Predicts measured substrate J/q within ~5%. 14
tests (+ slope). Validated live.
fox 2026-05-21: (1) use REAL API token usage, not len//4; (2) the
substrate prefills a large retrieved CONTEXT as INPUT while solo feeds
~nothing, so per-completion-token over-charges the substrate — and per-
TOTAL-token UNDER-charges it (its mix is ~98% cheap prefill tokens).
Measured n=30 qwen-nothink/4090: substrate prefills ~6.6k input tok/query
(claim_lattice) vs solo ~52 — ~127x. Neither single per-token denominator
is honest; prefill (parallel, cheap/tok) and decode (autoregressive,
dear/tok) must be costed separately.
- OpenAICompatibleClient stashes data['usage'] as .last_usage (non-
invasive; return type unchanged).
- watt_bench captures real prompt_tokens + completion_tokens per call
(both arms), aggregates per cell, and energy_cogs reports gross +
marginal per BOTH 1k-total-tok and 1k-completion-tok plus the context
size. Prints the prompt/completion split.
- 12 tests incl. the prompt-context artifact (per-total cheap, per-
completion dear). Full suite 2540 passed.
The clean per-input-tok / per-output-tok split rides bench/watt_calibrate
(slope calibration; separate commit once validated live).
fox 2026-05-21: 'we dont touch card during retrieval'. Retrieval + verify
are CPU/SQLite on the orchestrator; the GPU is idle (at the shared, always-
on model-resident floor) during them. So the substrate does NOT 'hold the
card' through its wall-clock — the gross window integral over-counts by
charging that always-on floor for the seconds we spend retrieving (energy
that exists regardless of the query; the card serves other traffic then).
Correct attribution: the GPU cost of a query is its GENERATION energy only
(the marginal — burst above the serving floor). Relabel: marginal is the
headline GPU COGS; gross is demoted to a reference 'window total, not
query-attributable'. Print + energy_cogs docstring + stock-v1-config doc
updated. No math change (marginal was already right) — this corrects the
narration. Measured qwen-nothink/4090: substrate GPU COGS is LOW
(claim_lattice $0.21/M-tok, quote $0.83) — its real overhead is latency
(CPU retrieval), not GPU watts.
fox 2026-05-21: 'gen 200W' was a bug — joules/window blends the ~400W
generation bursts with the sub-100W gaps (retrieval/verify/network) into
a power state the card never sits at. A card occupies DISTINCT states
(idle / middle-idle = resident-between-requests / generation), differing
per card×model×server.
watt_probe.classify_power_bands(): largest-gap split of the window
samples into a low band (serving floor) and high band (generation draw)
+ duty cycle. Data-derived, never hardcoded — tested at two scales. The
worker emits the decomposition + raw samples; RemoteProbe/LocalProbe
expose band_stats() uniformly.
energy_cogs: marginal now taken against the measured SERVING FLOOR (the
standing cost of being ready), not deep idle; the blend is kept but
labelled window_mean_w. Reports idle/serving-floor/gen-draw/duty.
Cache-miss certainty (fox's question): the arborist arm runs
burn_existing=True (force-deletes any live providence row before
inference) and asserts cache_hits==0 with a loud warning + real_inference
flag — so we time real generation, never a SQLite lookup. Solo has no
cache path. 11 tests (energy math + band split). Validated live on the
isolated 4090: solo gen 308W/70%-duty vs substrate 396W/8.6%-duty —
substrate marginal/tok is LOWER, gross/tok higher (it holds the card
longer for retrieval).
fox 2026-05-21: compute cost-of-goods-sold by kWh vs tokens, with the
three power states (idle / warm-idle / generation) MEASURED per
card×model×server — never hardcoded (his 40/127/380 W were illustrative
of one 3090). The only operator input is --price-per-kwh (default 0.33
USD/kWh, a configurable site rate).
energy_cogs() (pure, unit-tested) decomposes measured generation energy
against the measured warm-idle baseline:
* gross — all measured joules over the window (all-in, includes the
warm-idle cost of keeping the model hot, amortized).
* marginal — joules ABOVE warm-idle: what one more request's burst
actually costs (clamped >=0).
kWh = J/3.6e6; $/1k-tok is the unit that compares to API pricing. Both
surface per cell + a COGS print line.
watt_bench's arborist arm now loads the frozen bench.stock_v1 policy
(--answer-mode, drift-guarded on non-reasoning) so cost is measured for
the SAME substrate the campaign grades. Cells record
window_start/end_unix so a post-hoc load_monitor queue-depth cross-ref
can flag organic-traffic contamination on the non-isolated single-slot
endpoints. 6 COGS tests; full suite 2534 passed.
fox 2026-05-21: characterize substrate-ON under BOTH answer shapes, so
answer_mode is a swept axis, not a single pinned value.
stock_v1.py now exposes STOCK_V1_POLICIES{quote,claim_lattice} +
STOCK_V1_GOVERNANCE_HASHES (quote 5b6ca4c5..., claim_lattice 036a4c79...),
policy_for(mode), and assert_not_drifted(mode). Shared pins (crosslang
OFF, repair OFF, quantifier dry-run, metacognition label-only,
soft-preflight OFF, claim cap 12, v2-acronym-aware) are frozen
identically across modes.
Wire the treatment arms to the pin (the consumer-side step that makes
the freeze real):
* control_ab --answer-mode {quote,claim_lattice}
* control_sweep --arborist-answer-mode {quote,claim_lattice}
Both default claim_lattice (prior behavior), call assert_not_drifted on
non-reasoning runs (halts the sweep if DEFAULT_QUERY_POLICY drifts), and
load the frozen policy_for(mode) instead of an inline
dict(DEFAULT_QUERY_POLICY, ...). Reasoning refs (phase 3) keep their
documented JSON overrides and skip the assert by design (different hash).
jaggedness is left standalone — it is a mode-agnostic retrieval
instrument, coupling it to the answer-policy freeze adds friction with no
correctness gain. Full suite 2528 passed.
Before the multi-day campaign (hermes 3090/4090 -> qwen 3090/4090 ->
reasoning variants) the substrate-ON treatment arm must NOT drift. It
previously inherited DEFAULT_QUERY_POLICY implicitly, so any mid-run
edit would silently change what 'substrate-ON' means.
bench/stock_v1.py snapshots DEFAULT_QUERY_POLICY + re-asserts the
load-bearing pins (answer_mode=quote, crosslang OFF, repair OFF,
quantifier caps dry-run, metacognition label-only, soft-preflight OFF,
claim ceiling 12, v2-acronym-aware), then hashes the whole effective
dict. assert_not_drifted() fails loudly if that hash ever changes —
re-pinning is a deliberate fox-gated V.2 bump, never silent. The whole
campaign is identified by one governance_policy_hash
(5b6ca4c5...aade4e). Non-reasoning + non-distributed are harness axes
(reasoning -> phase 3, mesh -> later fork), not policy fields.
docs/stock-v1-config.md documents V.1, substrate-OFF (control_ab arm A),
the campaign matrix, and the energy-COGS companion (#000057) — whose
power states (idle / warm-idle / generation) are MEASURED per
card+model+inference-server at runtime, never hardcoded; only $/kWh is
an operator flag.
Answers fox's 2026-05-21 question — 'are we being swamped because we're
open to internet?' — before the multi-day GPU bench, where uncontrolled
internet traffic on the single-slot hermes (3090/vLLM) and qwen
(4090/llama.cpp) endpoints would contaminate wattage + throughput.
Three stdlib subcommands (urllib + sqlite3 + hand-rolled SVG, nothing to
install):
poll — scrape each endpoint's Prometheus /metrics on an interval into
SQLite; queue depth (num_requests_waiting / requests_deferred)
is the swamp signal a single GPU slot exposes. Prunes past
--retention-days each cycle (bounded store, no cancer growth).
graph — multi-panel SVG: queue depth, running, req/s, tok/s, e2e latency.
access — parse Caddy/nginx access log for the real client IPs the backend
can't see behind the proxy hop; top talkers + per-IP rate SVG.
Backend /metrics = HOW MUCH; proxy log = WHO. make monitor-poll /
monitor-graph / monitor-access.
v1 of the same-model substrate-delta harness's non-jagged metric.
For one corpus title, surface-perturb its question (numeral / accent /
hyphen / honorific / amp / brit) preserving the referent, then ask
whether retrieval surfaces the SAME target for canonical vs perturbed
phrasing. J_norm = XOR disagreement rate @k (lower = less jagged);
graded mean |Δrank| catches rank instability the binary metric misses.
Pure query --dry-run: no LLM, no verifier, no judge, no n=3 noise, no
5pp floor — the recall_at_k discipline. Reuses recall_at_k.probe +
mine_questions._surface_variant. Feeds #000012 ForkScore
ΔJaggednessReduction. A-vs-C answer-quality arm already exists under
#000057 (control_ab/control_sweep) — not rebuilt. Curvature + LLM-arm
jaggedness delta remain open (ticket §8).
make bench-jaggedness JAGGED_LIMIT=40 JAGGED_K=8
fox: a graveyard that keeps every wrong answer forever grows unbounded
like cancer and is counter to constraint-optimization. Bake in the
bound:
- earn-to-enter (recurrence-gated) — a failure earns a tombstone only
when its canonical claim-family re-fails; one-offs dropped.
- fingerprints not transcripts (UTXO-set analogy) — store
canonical_claim_family|falsifier_root|failure_class, prune bulk text.
- decay/compact to steady-state — evicts like a surface, never a core.
- off the hot path — dedicated graveyard shard, bloom-filter burden
check only.
Gossip-group falsifier admission inherits the same discipline:
difficulty-adjusted stable-rate ingestion (BTC block-rate target) +
per-window budget (#000036), enforced in #000012/mesh. BTC lesson is
bounded self-regulating ingestion, not store-everything. If it cannot
be bounded to a steady state, the graveyard is not built.
Three tickets from the 2026-05-20 Dav1dPrometheus "Protocol-Layer AGI"
working report (held outside the repo; referenced not committed):
- #000058 cache_key_9 verifier-policy mandatory-vs-legible decision +
doc reconcile. Records the five-step-#1 correction: verifier fields
already fold into governance_policy_hash, so the 9th dim is audit
legibility not a correctness gap.
- #000059 admission discipline: claim-graveyard burden-shift +
self-providence quarantine (guards the existing ingest-self-providence
self-confirmation loop).
- #000060 H-ABCDEFG same-model substrate-delta harness (jaggedness +
curvature); curvature-aware ForkScore folded into #000012 Phase 2,
not spawned as a sibling.
Reconciles CLAUDE.md cache_key invariant (8-dim -> 8 + optional 9th).
Next ID 000058 -> 000061.
Implements fox's 2026-05-20 architecture: don't ship shards to the
tight-on-disk GPU boxes; run the benchmark FROM the laptop (local
shards + retrieval + judge + workload loop, driving the worker's LLM
endpoint over the network) and have the worker boxes REPORT their own
power.
Adds a probe abstraction with two implementations behind one
start/stop/gpu_stats/cpu_stats interface:
- LocalProbe: wraps PowerSampler + CpuSampler (watt_bench runs ON the
GPU box — the original mode)
- RemoteProbe: orchestrates bench/watt_probe.py on a remote worker
over SSH (scp the stdlib probe once, launch --until-file detached,
touch the stop-file after the workload, fetch the JSON). The worker
needs no shards / arborist / venv.
--remote-gpu-host HOST selects RemoteProbe; --endpoint points at the
worker's endpoint. Idle baseline now records gpu_util_mean_pct and
flags ">5% ⇒ contaminated by live traffic" so the 3090-style
contamination (live hermes traffic polluting the baseline) is visible
in the output, not silent.
Verified end-to-end: laptop-driven n=2 solo qwen, power sampled on the
4090 over SSH — idle gpu 20.2W @ 0% util (clean) + cpu 86W; load
populated; report tagged mode=remote. The on-box LocalProbe path is
unchanged (omit --remote-gpu-host).
Report schema gains mode / remote_gpu_host / gpu_available /
cpu_energy_available. Per-cell records gpu+cpu+total joules/question.
Splits the power-sampling half out for the laptop-driver / worker-
reporter architecture (fox 2026-05-20): don't ship shards to the tight-
on-disk GPU boxes; instead drive the benchmark FROM the laptop (which
holds the shards + judge + workload loop) and have the worker boxes
just serve the model and REPORT their own power.
Power sampling must run on the box (nvidia-smi + RAPL are host-local),
but nothing else does. watt_probe.py is stdlib-only — python3 +
nvidia-smi + readable RAPL, no venv, no arborist, no shards. Copy it to
the worker and run; it samples GPU power (nvidia-smi, trapezoid-
integrated to joules) + CPU package energy (RAPL energy-diff) over a
window and emits the same energy schema watt_bench's local samplers
produce. Two window modes: --duration (fixed) or --until-file (laptop
touches a stop-file — the workload-correlated mode). Records
gpu_util_mean_pct so live-traffic contamination can be detected.
Takes only the top-level RAPL package (intel-rapl:N, single colon),
not the :N:M core/uncore subdomains, so no double-count.
Verified on the 4090: scp'd to /tmp, ran with no install, real readings
(GPU 20.2W idle @ 0% util, CPU 82W, joules computed, 6 samples / 3.2s).
Next: laptop-side orchestration in watt_bench (--remote-gpu-host) that
SSH-starts this probe for each cell's window while the driver runs
retrieval + judge locally and drives the worker's LLM endpoint.
The exact llama.cpp (qwen/4090) and vLLM (hermes/3090) launch commands,
recon'd 2026-05-20, so the foxhop-states salt states can be written
accurately rather than guessed. Notes the convention (salt manages the
systemd unit; engine binaries + model artifacts stay manual on
/mnt/data as documented prereqs) and the live-hermes cutover constraint
(keep >=1 hermes online; qwen is expendable).
fox's call: capture the RAPL permission as a Makefile target run with
sudo, not an ad-hoc chmod. `sudo make rapl-access` on each GPU box
installs a udev rule (/etc/udev/rules.d/99-rapl-readable.rules) that
makes intel-rapl energy_uj world-readable on every powercap add event
(survives reboot), and applies chmod immediately so no reboot is
needed. Idempotent; reversible via `sudo make rapl-access-revoke`.
energy_uj is root-only by default (PLATYPUS side-channel mitigation,
CVE-2020-8694) — that's why bench/watt_bench.py's CpuSampler read null
CPU watts as fox during recon. After this target runs, watt_bench
reads CPU package energy directly (no --cpu-energy-cmd needed). GPU
watts via nvidia-smi never needed special perm.
Non-root guard + help entries verified; Makefile parses clean.
Toward fox's next goal: score the full serving stack on quality AND
cost — {qwen, hermes} × {llama.cpp, vLLM} × {3090, 4090} × {solo,
arborist}, measuring CG% + GPU watts + CPU watts + joules/answer per
cell.
watt_bench.py — adds CpuSampler (Intel RAPL package energy via
/sys/class/powercap/intel-rapl:*/energy_uj). RAPL exposes a cumulative
microjoule counter, so energy-over-window is an end-minus-start diff
(handles wrap) — more accurate than integrating instantaneous power.
Sums multi-package. energy_uj is root-only by default (PLATYPUS /
CVE-2020-8694), so it degrades to available=False when locked;
--cpu-energy-cmd 'sudo cat {path}' supplies a privileged reader when a
sudo rule exists. Each cell now reports gpu/cpu/total joules-per-
question + gpu joules-per-token; the report records cpu_rapl_available.
Verified: graceful degradation when locked; RAPL diff math (1->4 MJ uJ
= 3.0 J, exact).
benchmark-matrix.md — expands the cost section to the full 16-cell
(model × engine × GPU × arm) design, the per-cell metric set (quality +
GPU + CPU energy), the serving-stack inventory from 2026-05-20 recon
(4090=qwen/llama.cpp, 3090=hermes/vLLM — each box has one engine + one
model today), and the buildout gap (vLLM+qwen, llama.cpp+hermes, cross-
GPU models). Notes idle-floor asymmetry (hermes/3090 ~127W vs
qwen/4090 ~20W) as a real optimizer input.
Harness is ready; the serving-config buildout + RAPL perm grant are the
remaining (ops, fox-directed) prerequisites to run the full matrix.
Two deliverables for the cost/energy axis of the constraint optimizer.
docs/benchmark-matrix.md — shareable spec of the control experiment:
the question, fixture (386 office-holder Qs with corpus-vintage gold),
the 3-model × 3-framing × 2-arm matrix (18 cells), the verdict
vocabulary + two reads (accuracy vs grounding-fidelity), the
deterministic code judge + its Opus calibration, the results-so-far
table, and the NEW cost dimension (tokens / latency / GPU watts /
joules-per-answer measured per GPU tier). Self-contained — readable
cold by David.
bench/watt_bench.py — GPU wattage harness. Samples nvidia-smi
power.draw on the inference GPU while driving a small representative
subset, reports mean/peak watts, trapezoid-integrated joules,
joules-per-question, and joules-per-token. Tags the GPU
(--gpu-label 3090|4090) so the optimizer can compare hardware tiers.
Idle-baseline sampling separates load draw from idle. Does NOT grade
(energy is independent of correctness); saves answers + per-question
timing to JSONL for a later quality-per-joule pass via
score_with_code_judge.
Designed to run ON the GPU box (the orchestrator has no GPU; the
3090/4090 live on the inference boxes). Degrades gracefully when
nvidia-smi is absent (energy fields null) so it is testable anywhere.
Verified: PowerSampler graceful degradation + trapezoid integration
(synthetic 100->200->200W over 2s = 350 J, exact).
The headline cost finding the optimizer must weight: qwen-think
reasoning = 1300-3300 tokens/answer vs qwen-nothink ~50-100 (20-50x),
for a workload where arborist+qwen-nothink already lands 82% CG. The
energy numbers will quantify whether reasoning's premium is ever
justified — grounding-fidelity per joule, not per answer.
fox was right ("or you have defects still"): the arborist+qwen-think
empties were NOT an inherent reasoning limitation, they were
max_tokens set too low. Diagnosed with finish_reason + token usage on
a realistic large arborist-style context:
max_tokens=1024 → 4/4 empty, ALL finish_reason='length', all hit
exactly 1024 tokens
max_tokens=4096 → 0/4 empty, ALL finish_reason='stop', used
1339-3295 completion tokens
qwen-think spends 1300-3300 tokens on its (internal) reasoning trace
BEFORE emitting the tiny claim-lattice JSON. The arborist arm's budget
comes from DEFAULT_QUERY_POLICY["max_tokens"]=512 (correct for non-
reasoning single-line JSON), so reasoning refs hit finish='length'
mid-trace and return EMPTY. Fix: arb_policy["max_tokens"]=8192 for
reasoning refs (generous headroom over the 3295 observed ceiling).
This also CORRECTS a wrong claim in commit aa9d9c8's message: the
json-schema grammar does NOT suppress the reasoning trace. The
reasoning happens (it burns those 1300-3300 completion tokens); the
grammar only shapes the final emitted answer into schema-valid JSON.
So arborist+qwen-think is a genuinely distinct config, not a clone of
arborist+qwen-nothink. The earlier "grammar suppresses reasoning"
read was an artefact of toy-prompt diagnostics where reasoning fit
under 1024 tokens.
The empty-retry from aa9d9c8 stays as belt-and-suspenders for any
residual model-side empties, but with an adequate budget it should
rarely fire.
Cost note (fox's standing point, now quantified): reasoning = 1300-
3300 tokens/answer vs qwen-nothink's ~50-100 = 20-50x token cost. For
a workload where arborist+qwen-nothink already lands 82% CG, that cost
multiplier is the open question — not whether the cell is measurable
(it now is), but whether the marginal lift justifies 20-50x spend.
Two defects blocked the arborist+qwen-think cell, both diagnosed
2026-05-20:
Defect 1 — stop sequence truncates to empty. The claim_lattice path
sets stop=['\n\n'] (runaway guard tuned for single-line Hermes JSON).
A reasoning model's output trips it immediately → 100% empty answers
→ 100% ABSTAINED (measured on the first 23 items of the killed run).
Fix: MODELS gains an explicit flag; the arborist arm
clears claim_lattice_json_stop_sequences for reasoning refs. Direct
A/B confirmed: stop=['\n\n'] → ''; stop=None → valid JSON.
Defect 2 — intermittent empty completions. Even with the stop cleared,
qwen-think under json-schema grammar emits an empty completion ~1/3 of
calls (a llama.cpp reasoning+grammar artefact; qwen-nothink phase 3
had ~0 spurious empties). Fix: empty-output self-heal — reasoning refs
retry up to 3 attempts, burning the cached empty each retry. Never
fabricates: a still-empty answer after retries is recorded as empty.
6-item smoke: 0/6 (broken) → 5/6 valid JSON; residual ~1/6 are
questions that reliably break (4 consecutive empties), a documented
artefact.
Structural finding (to fold into Addendum 8): json-schema grammar
enforcement SUPPRESSES the reasoning trace — output is pure single-
line JSON, no <think> block. So arborist+qwen-think is structurally
≈ arborist+qwen-nothink; the thinking lever that moved the solo arm
is neutralized by the claim_lattice grammar. The re-run will confirm
empirically.
Non-reasoning refs (hermes, qwen-nothink) unchanged: single pass,
stop sequence intact.
Durable record of the #000057 sweep in the bench journal. Captures:
- The question: is Hermes-8B's confident present-day-officeholder
fabrication an 8B weakness, a framing artefact, or does retrieval
fix it? Crosses {hermes, qwen-nothink, qwen-think} × {plain,
source_relative, as_of_corpus} × {solo, arborist} on a 386-item
office-holder fixture with corpus-vintage gold.
- The judge methodology: Opus headless judge burned quota (79.5%
JUDGE_ERROR), replaced with the deterministic code judge
(bench/judge_code.py), calibrated against Opus's gradeable records
(CG agreement 13->47%, WRONG 56->89%, ABSTAINED 80->95%).
- Consolidated CG% scorecard, all arms on the identical final judge.
- Three findings:
1. Retrieval dominates — arb/qwen-nothink/plain 82% vs 7% solo;
no solo config approaches the retrieval arms.
2. Reasoning does NOT improve raw correctness — qwen-think/as_of
44% vs nothink 50%.
3. Reasoning's real cost is broken honest-abstention —
qwen-nothink/source_relative abstains 97% (clean); qwen-think
only 61%, reasoning itself into wrong parametric answers.
- Production recommendation: arborist + qwen-nothink, plain framing,
reasoning OFF (82% CG, ~0% abstain, 11% wrong-assert).
- Held cell noted: arborist+qwen-think running at write time, result
to be appended.
Bench %s are point-in-time measurements (not repo-derived counts),
so no AUTOCOUNT tags — consistent with addenda 1-7. test_doc_counts
3/3.
fox 2026-05-19: 'we don't need to redo anything'. The arborist+qwen-
nothink sweep would re-run qwen-nothink solo on 1158 records that
phase 1 already produced (control_sweep_2026-05-19T21-52-56Z.jsonl,
calibrated rescore on disk). Wasteful — qwen-nothink solo behaviour
doesn't depend on retrieval being on, so phase 1's numbers stand.
--skip-solo runs ONLY the arborist arm. Implementation: skip the
per-variant model loop in _process_item when the flag is set; the
arborist arm below it still runs if arborist_on. The spend banner
zeroes the solo-call count so the operator sees the actual LLM
budget for the arborist-only run.
Example: a full arborist+qwen-nothink sweep against the 386-item
fixture goes from 2316 LLM calls (1158 solo + 1158 arborist) to
1158 calls (arborist only). Halves wall-clock on single-worker
llama.cpp.
pytest 27/27, imports clean, --help shows the flag.
Reorder rule 3 (short-answer entity grounding) above rule 4 (NLI
contradiction) so positive lexical evidence cannot be overridden by
NLI clause-level noise. Surfaced by the 2026-05-19 arborist+qwen-
nothink smoke:
i=3 · who is the prime minister of Poland?
ans: 'Donald Tusk is listed as the Prime Minister of Poland.'
gold: ...lists Tusk + Marcinkiewicz + Belka + Kaczynski + Kopacz...
NLI contradiction p=0.892 (above 0.85 threshold)
NLI entailment p=0.744 (also high on the correct clause)
Tusk WAS PM in 2010 (served 2007-2014); answer is correct against
the corpus-vintage gold. The NLI contradiction signal came from
clause-level candidate selection picking a NON-Tusk PM the source
also mentions; entailment was high on the Tusk clause. Mixed signal
that the WRONG rule then over-confidently resolved.
The fix is a rule reorder, not a threshold change — the fast path's
positive-evidence combination (specifics-in-gold AND subject-in-gold)
is a strictly stronger signal than NLI's clause-level max
contradiction, so when it fires it should win. The combination
discriminates Poland-Tusk (Tusk ∈ gold, Poland ∈ gold → CG) from
Anthony-Albanese (Albanese ∉ gold → fast path declines → falls
through to UNGROUNDED-subject-in-gold → WRONG, unchanged).
Self-test 4/4 INSTRUMENT TRUSTWORTHY unchanged. pytest 27/27.
Poland-Tusk regression smoke: now CG via short_entity_grounded ✓.
No regression risk on the existing reconciliation cells:
- Iceland CG: short_entity_grounded was already winning (was rule
4, now rule 3 — same outcome, earlier exit)
- WWII-1812 WRONG: '1812' ∉ gold → fast path declines, NLI fires ✓
- Higgs-cafe FABRICATED: 'Higgs' ∉ gold → fast path declines ✓
- Anthony Albanese WRONG: 'Albanese' ∉ gold → fast path declines ✓
- Abstention phrases: rule 2 still fires first ✓
Two surgical fixes unblock 'arborist with synthesis LLM = Qwen-on-
llama.cpp' as a viable arm in the control sweep. Pre-existing
docstring said 'Arborist×Qwen needs proof-path guided_json+extra_body
surgery — coupled follow-up'; this is that follow-up.
Fix 1 — multi-engine structured-output extras
The runner / query JSON-mode paths previously sent only vLLM's
'guided_json' key for the claim_lattice schema. llama.cpp silently
drops it, leaving Qwen un-enforced (the parse-tolerant fallback did
all the work). Helper
claim_lattice_structured_output_extras() in arborist/qa/verify.py
now returns a dict carrying the schema under all three engine
conventions:
- guided_json (vLLM grammar-constrained sampling)
- json_schema (llama.cpp native shorthand)
- response_format (OpenAI-spec, honoured by llama.cpp and newer vLLM)
Each engine recognises its own key and silently drops the others.
Used at both inference call sites (runner.py:740, query.py:3324).
Hermes/vLLM path is unchanged — it picks up 'guided_json' and
ignores the other two.
Fix 2 — query() accepts user-supplied extra_body, merges with defaults
query() grew a keyword-only extra_body parameter (default None).
Per-model knobs (Qwen's {'chat_template_kwargs': {'enable_thinking':
False}} toggle, future template knobs) can flow from the caller to
the synthesis chat-completion call. Schema-enforcement extras are
added inside query() and merge under user keys — common case is
disjoint namespaces, but if a caller wants to override 'guided_json'
they can.
bench/control_sweep.py now passes MODELS[arborist_ref]['extra']
through to query() in the arborist branch, so --arborist-ref
qwen-nothink runs with reasoning disabled and --arborist-ref
qwen-think runs with reasoning enabled. Phase 1's arborist arm with
--arborist-ref=hermes is unaffected (MODELS['hermes']['extra'] is
None, merges to no-op).
Tests
+ 3 new in tests/test_verify_json.py covering helper default shape,
alternate-schema reuse, and query()'s new extra_body parameter
220 affected tests still green (verify / claim_lattice / judge /
runner suite)
pytest test_verify_json: 27/27
Next: small smoke run --arborist-ref qwen-nothink against 4-8 items
to confirm end-to-end before any full sweep. Phase 2 (qwen-think solo)
still running in background, unaffected — it doesn't touch the
arborist arm.
The Arborist arm runs answer_mode='claim_lattice' (per control_sweep.py
:179, control_ab.py:155) so its answers arrive as the JSON envelope
{"claims":[{"text":"...","evidence_ids":["E1"]},...]}.
_descaffold strips the [E1] evidence-pointer markup but the JSON
braces + key syntax remain. The verifier's strategy-2 (span) and
strategy-3 (proper-noun) extractors see brace noise instead of the
inner claim prose — every Arborist record degraded to UNGROUNDED.
The 2026-05-19T17-01-17Z sweep, re-graded with the freshly calibrated
judge (5a17f61), surfaced this: Arborist arm reported 0 CG across all
three variants in the live phase 1 output (the live run was pre-
calibration), and 29/120 CG (24%) under the calibrated rescore — clear
improvement just from theta_contra=0.85, but the JSON envelope was
still hobbling the verifier paths.
Fix: _unwrap_claim_lattice_json runs BEFORE all downstream rules.
Detection is conservative (three independent signals: starts-with-
brace AND "claims" key AND "text" key) so plain-prose answers
pass through unchanged. Multi-claim envelopes concatenate as discrete
sentences (extract_claim_spans treats each as its own span).
Malformed JSON falls back to the original answer — no silent
rewriting on broken input.
Smoke result on the Iceland Arborist case
ans: {"claims":[{"text":"The current president of Iceland is
Ólafur Ragnar Grímsson.","evidence_ids":["E1"]}]}
gold: {{Infobox Political post |post = President |body = Iceland
|incumbent = [[Ólafur Ragnar Grímsson]] ...}}
before: UNGROUNDED → FABRICATED (then WRONG after calibration)
after: short_entity_grounded → CORRECT_GROUNDED
pytest: 27/27 (added 7 unwrap-coverage tests covering single-claim
envelopes, multi-claim concatenation, plain-prose passthrough,
malformed-JSON tolerance, unrelated-JSON passthrough, and the
end-to-end Arborist-envelope CG flow). Self-test 4/4 unchanged.
Re-rescores of 17:01 sweep + phase 1 sweep run after this commit
to measure final Arborist scorecard improvement.
Opus is trusted; this commit closes the systematic gaps surfaced by
B's rescore on the 17:01 sweep (2289 records, of which 468 had real
Opus verdicts). Off-diagonal cells in descending size and the
root-cause fix for each:
cell n=114 code:WRONG · opus:CORRECT_GROUNDED
Root cause: NLI fires contradiction p in [0.5, 0.75] on factual
answers like 'Ólafur Ragnar Grímsson is president of Iceland'
against wikitext-shaped infobox gold — clause-level candidate
selection picks up co-mentioned earlier office-holders, reads
temporal-frame mismatch as contradiction. The TRUE contradiction
signal (WW2 1812 self-test fixture) measures p=0.985 — clean
margin above noise.
Fix: raise theta_contra 0.5 → 0.85 (code-judge override of the
NLI manifest's 0.5 contradiction_veto).
cell n=60 code:FABRICATED · opus:WRONG
Root cause: 'Anthony Albanese' answer vs Julia Gillard gold gets
FABRICATED (specifics-not-in-gold) but Opus correctly distinguishes
WRONG (source contradicts by naming someone else) from
FABRICATED (source silent on topic).
Fix: when verifier UNGROUNDED + specifics not in gold AND the
question's subject anchor IS in gold, demote FABRICATED → WRONG.
Subject anchor uses proper-noun-shaped terms from the question
(Iceland / Australia / Higgs) — not the last-content-token
heuristic, which mis-fires on coincidental matches like 'cafe'
appearing in a 'gold does not mention any cafe' denial.
cell n=18 code:WRONG · opus:ABSTAINED
Root cause: abstention patterns missed Hermes's most common
refusal phrasings — 'I do not have accurate information', 'I do
not have access to a reference knowledge base', 'I lack access
to'. Original patterns required determine/know/tell verbs right
after 'do not'.
Fix: three new patterns for the 'do not have / lack ...
information / access / knowledge' family.
cell n=13 code:ABSTAINED · opus:CORRECT_GROUNDED
Root cause: verifier's strategy-2 needs prose shape; terse-name
answers ('Pratibha Patil', 'Jalal Talabani') fall to
UNGROUNDED-no-specifics → ABSTAINED, missing valid CG.
Fix: short-answer entity-grounding fast path. When answer is
short (≤15 tokens) AND every specific asserted is present in
gold (no unsourced) AND at least one specific WAS asserted AND
the question's subject anchor is in gold → CG. Guards against
'wrong topic, right name' false-positives via the subject check.
Structural reorder: NLI contradiction now runs AFTER the
abstention check and short-answer fast path (instead of preempting
the verifier), so the verifier's STRICT/HYBRID positive signal
isn't overridden by NLI noise. NLI still leads the path on truly
unbounded answers — verifier UNGROUNDED + NLI ≥ 0.85 contradiction
keeps the WRONG label.
Self-test 4/4 INSTRUMENT TRUSTWORTHY. pytest contract 18/18.
v2 rescore on the same 17:01 sweep runs in the background to
measure agreement-matrix improvement empirically.
Added bench/analyze_judge_disagreement.py — the harness that drove
this calibration (reads B's rescore JSONL, bucketises off-diagonal
cells, dumps configurable samples per cell with question / answer /
gold / both rationales). Reusable for the next calibration round.
Reads an existing sweep JSONL (Opus verdicts already recorded), re-fetches
gold per record via _gold(), runs the code judge on (question_asked,
answer, gold), and emits:
- markdown scorecard: agreement matrix (code × original judge), per-arm
/ per-model / per-variant code-judge tallies, and a residue table of
the JUDGE_ERROR records (the natural input to a later LLM-batch
needle-haystack pass — Opus or Grok);
- JSONL with one row per sweep record (code_verdict + code_rationale
+ code_decision), joinable on (i, arm, model, variant) to the
source sweep.
Zero LLM calls. Reads sweep JSONL + shards read-only. Pairs with the
new --judge switch (a2e9b49): the switch decides what NEW data uses;
this script decides what the ALREADY-COLLECTED data looks like under
the deterministic judge.
Usage (parameter default matches control_sweep.py default fixture):
python -m bench.score_with_code_judge --in <sweep>.jsonl
Currently running against control_sweep_2026-05-19T17-01-17Z.jsonl
(the 2289-record sweep that ran on the prior huge-N pass before the
Opus quota burned out). Output will land at
bench/qa_results/control_sweep_2026-05-19T17-01-17Z_code_judge.{md,jsonl}.
Wire bench/judge_code.py into the sweep harnesses as the default judge.
Both control_sweep.py and control_ab.py grow a --judge {code,opus} CLI
arg; both share the same Verdict shape so the dispatch is a pointer
assignment + threading the judge_fn through _process_item.
Behaviour:
- DEFAULT = code: zero LLM, zero quota, deterministic. Self-test gate
is the code judge's 4-fixture contract.
- --judge opus: original gated Opus path; needs ARBORIST_JUDGE_ENABLE=1
set per 1cabfe6's fail-closed gate, otherwise every record returns
JUDGE_ERROR with rationale 'disabled — set ARBORIST_JUDGE_ENABLE=1'
and the sweep records that label honestly.
Reporting:
- Header line now records which judge ran ('Judge = code (...)' or
'Judge = opus (...)') so partial-reports & resumes don't lie about
provenance.
- Spend banner shows '0 LLM calls' for the code path so the no-burn
property is visible in the operator output.
Test surface: pytest sweep across tests/ still 136/136 (no regressions);
new --judge flag visible in --help on both harnesses.
Next: bench/score_with_code_judge.py to re-grade existing sweep JSONLs
(written under the gated-Opus run) with the code judge; agreement
matrix surfaces residue size for the eventual LLM-batch needle-haystack.
bench/judge_code.py — drop-in alternative to bench/judge.py with the
same Verdict shape & closed verdict vocabulary (CG/W/F/A/JE) but zero
quota cost: composes verifier + NLI + abstention + specificity into a
fixed-order pipeline. fox 2026-05-19: 'data first, judging later' —
this is the data-collection arm; LLM-based judging (Opus batched
needle-haystack, or Grok credit-card) is a separate downstream
concern that operates on the residue this judge cannot classify
deterministically.
Pipeline (first hit decides):
1. empty / no-gold guards
2. explicit abstention phrases (lexical regex)
3. NLI contradiction (arborist.qa.nli.shadow_check) — strongest
signal: gold contradicts the claim → WRONG
4. lexical verifier (arborist.qa.verify.verify_quotes) →
STRICT → CORRECT_GROUNDED
HYBRID + NLI entail >= 0.55 → CORRECT_GROUNDED
UNGROUNDED + specifics-not-in-gold → FABRICATED
UNGROUNDED + no specifics → ABSTAINED
HYBRID without NLI corroboration → JUDGE_ERROR (residue
for an LLM judge)
Threshold note: _CODE_JUDGE_THETA_ENTAIL_CORROBORATE=0.55 is distinct
from the NLI manifest's entailment_block_veto=0.9. The manifest's
threshold is calibrated for OVERRIDING a STRICT lexical signal with
negative evidence — high bar. The corroboration use here is the
opposite direction: additive positive evidence on an already-positive
anchor — moderate bar appropriate. Self-test case 1 measures NLI
entail=0.769 (clearly entailed, clear margin above 0.55).
Specificity for FABRICATED layers three scanners:
- verifier's multi-word proper-noun extractor (Higgs Boson, ...)
- local single-word capitalised-token scanner (Napoleon, Mars, ...)
deliberately separate because the verifier's gate is conservative
by design (multi-word only)
- numerics (years, dates, large counts, money)
Self-test: same 4 fixtures as bench/judge.py:self_test() so the two
instruments can be cross-checked when fox re-fires the Opus judge on
the residue later. Result: 4/4 INSTRUMENT TRUSTWORTHY.
tests/test_judge_code.py — pulls the contract into make test
(18 cases): module identifiers pinned, dataclass shape parity,
empty / no-gold guards, parametrised abstention phrases, specificity
layer behaviour, the canonical 4-case self-test, batch helper, and
graceful NLI-unavailable degradation. 18/18 pass.
Pre-existing known limitation, documented in the docstring: terse
correct answers ('In 1945.' against gold containing '1945') route to
ABSTAINED because the verifier's span extractor needs prose shape;
NLI sees no clause-level overlap at very short claims. The conservative
ABSTAINED label is correct deferral; tuning this is a calibration
question for real bench data, not the instrument's contract.
No callers touched yet — control_sweep.py & control_ab.py still
import the disabled Opus judge. Wiring this in is a separate ticket
move per fox's data-first sequencing.
2026-05-19: huge-N #000057 control sweep (f63b00d → 9dc02e4) burned
our Opus quota. Disable judge.py by default so a stray re-run can't
re-burn — every call short-circuits to JUDGE_ERROR with rationale
'disabled — set ARBORIST_JUDGE_ENABLE=1 ...' and zero subprocess
spawn (0ms in the disabled path, smoke-tested).
Why a gate, not a model swap:
- judge.py uses Opus deliberately as EXTERNAL SOTA outside both arms;
swapping the judge to Hermes/Qwen would corrupt the experiment
(Hermes is itself an arm under test). The hygiene comment at
judge.py:26-29 already names same-family-judging as the live threat
to validity at Opus level; downgrading further changes the science.
- Gating instead preserves the science when fox re-enables, and gives
us the data-first workflow he asked for: deterministic tool
pre-filters (verifier / NLI / recall@k) up front, judge only on
residue worth Opus tokens, with explicit go.
Behaviour:
- control_sweep.py + control_ab.py already treat JUDGE_ERROR
non-fatally (counted as JE in _bucket); disabled runs degrade to
100% JE in the tally and surface the disable reason in rationale —
the loudest possible 'judge did not run here' signal.
- Re-enable per-run: ARBORIST_JUDGE_ENABLE=1 python -m bench.control_sweep ...
- self_test() will report 4× JUDGE_ERROR when gated — intentional;
if the instrument is off, the self-test must NOT silently pass.
Smoke-test (without flag): label='JUDGE_ERROR' rationale='disabled — ...'
dt=0.0ms · no claude subprocess spawned.
Cross-referenced from CLAUDE.md '## Live endpoints' /
'Budget discipline' subsection added in 2365bd1.
2026-05-19: huge-N #000057 control-arm sweep (f63b00d → 9dc02e4) burned
our Opus quota. Pin the policy into repo CLAUDE.md so future agents
don't reach for Opus on autopilot:
- Sweep + bench-qa default endpoints: Hermes-3-8B + Qwen via uncloseai
- Opus drops out of control-arm grids until fox explicitly re-adds it
- Existing bench/results/ data stays valid; just no new Opus expansion
- Translation still uses local pinned opus-mt (sandwich MT edges), not
Hermes/Qwen 8B
- Grok rejoin path noted as an HTML comment (paid spend, fox-gated)
Adds two subsections under '## Live endpoints':
- 'Budget discipline — Hermes / Qwen first, Opus deferred'
- 'Per-call model selection (general agent work, not sweep arms)'
fox: 'not 11 hours it shouldn't take that long'. Probed endpoints —
Qwen-27B absorbs 8 concurrent with 0 errors (0.5->2.1s); the
bottleneck is the serialized claude -p Opus judge, so more workers
≈ near-linear speedup. control_sweep.py gains --resume PATH: appends
to an existing JSONL, skips items already COMPLETE (full
models×variants for solo + variants for arborist if i<=arborist-n),
re-runs partial items; _aggregate now dedupes (i,arm,model,variant)
last-wins so a killed-mid-unit restart never double-counts, and
_load_recs tolerates a truncated trailing line from the kill.
Makefile control-sweep gains CONTROL_SWEEP_WORKERS / _RESUME / _ARB_N
so make stays the interface. 6-worker run killed cleanly (specific
pids, no pkill), relaunched resume @ 12 workers — 12 done items
preserved, 374 to run, ~5h -> ~2-2.5h.
fox: 'make the n huge huge, check in every ~7 turns'.
- stale fixture re-mined to the FULL pool: 386 questions (180 pres /
119 PM / 34 premier / 20 gov-gen / 17 chancellor / 15 CM / 1 FM),
deterministic.
- control_sweep.py rebuilt: ThreadPoolExecutor work-unit-per-item
(per-unit qa_db -> zero SQLite write contention; shards read-only
safe under concurrent readers), incremental flushed JSONL, and a
--report-only PATH mode that aggregates a PARTIAL file with ZERO
LLM/judge spend (the interim check-in path).
- huge N goes on the CONTROL (solo x3 models x3 framings) where the
open statistical question lives; Arborist A/B = fixed --arborist-n
(default 40), not re-measured 386x (power belongs on the control,
and query() over ~40GB shards is the heaviest call).
- bench-maxing doctrine applied: independent hermetic judge calls
fanned out, 'serial-by-caution is halting in disguise'.
N=3 (prior run) already shows the coherent, review-relevant story:
plain -> all 3 models confidently assert (0% abstain); source_relative
-> abstain jumps (hermes 33 / qwen-think 67 / qwen-nothink 100%);
as_of_corpus -> ~100% correct all models. The gap is largely
framing+snapshot, not a universal capability deficit.
fox ruling: 'we both do not know which framing is right, measure all
benchmarks and bring results forward for review'; 'we have qwen with
and without reasoning to use'.
bench/control_sweep.py: sweeps the CONTROL (solo) arm across
{Hermes-3-8B, Qwen3.6-27B reasoning, Qwen3.6-27B no-reasoning} ×
{plain, source-relative, as-of-corpus-era} questions, judged vs the
fixed corpus-vintage gold; Arborist-Hermes treatment reference run
alongside at the same N. Presents the SAME judge verdicts under both
the accuracy framing (the naive read fox flagged as unfair-as-truth,
shown for contrast) and the grounding-fidelity framing (the
defensible read), plus an explicit note on the faithfulness-ablation
framing + Arborist×Qwen — both deliberately NOT run (proof-path
surgery / different instrument, not a relabelling).
Answers fox's open question empirically: does a 27B *reasoning*
control honestly ABSTAIN where the 8B fabricates (gap = weak-small-
model artefact) or still confidently assert post-corpus (gap = real,
scale-independent)? as_of_corpus separates 'can't recall the era'
from 'won't constrain to a source'.
Qwen toggles probed live: reasoning answer in message.content (CoT in
separate reasoning_content, not surfaced — logged limitation);
no-reasoning via chat_template_kwargs{enable_thinking:false}. Gate =
in-script judge self_test (aborts on fail) — stronger than a make
edge, no double self-test spend.
Opt-in third NLI inference backend (ARBORIST_NLI_BACKEND=tinygrad) in
qa/nli/shadow.py, parallel to torch/onnx-int8, behind the #000049
cage: shadow-only, never an audit_mode input, never auto-preempts the
proven path (guarded so it cannot regress torch/onnx). Loads the ONNX
export through tinygrad's frontend wrapped to the existing
model(**enc).logits contract so _nli_batch is byte-unchanged.
bench/scripts/nli_backend_ab.py + make bench-nli-backends: deterministic
A/B, gate is numerical agreement with the torch reference first,
latency second (a divergent engine = a different shadow signal = a
different nli_policy_hash). Instrument is honest — reports a requested
backend as unavailable rather than relabelling a fallback's numbers.
First CPU-smoke run already quantified that the deployed §7 #22 int8
export diverges Δmax≈0.42 from torch — the immunity property made
measurable, not a defect. Real tinygrad numbers pending a producer-box
run (tinygrad not an arborist dep; frontend op-coverage for the large
MNLI checkpoints unverified by design).
docs/onnx-vendor-capture-immunity.md: why the model-in-proof-path cage
makes the inference engine an interchangeable sidecar, never a trust
dependency — public-domain positioning capital. Indexed in CLAUDE.md.
Full suite 2498 passed (identical to baseline); 24/24 NLI tests green.
control_ab.py called client.chat_completion(messages) but model is a
keyword-only REQUIRED arg → TypeError every call → solo answer was
always '[solo-error: TypeError]', which the hermetic judge correctly
classified ABSTAINED. Every prior 'Hermes-solo abstained' smoke
(incl. the n=2 the ticket cites as a case against the treatment) was
a broken-control-arm artefact, not a measurement. Fix: pass
model=a.model (symmetry with the Arborist arm). Error token now
carries the exception message so the next failure self-describes.
Also adds the #000057 'stale' fabrication-bait mining class
(time-varying executive offices; corpus = ~2010-2011 Wikipedia
VERIFIED from the artifact, NOT the 2003 dump CLAUDE.md names —
flagged) and restores the numeral fixture deterministically.
b573c59's accent-fold work removed 210 lines from
qa_questions_numeral_map.json (40 -> 5 entries) — collateral of an
earlier `mine_questions.py --cls numeral --limit 5` regression-check
that got committed. The control-ab run scored n=5 (not the --n 12
passed) because the fixture itself only had 5 rows. Re-mined at the
default --limit 40; deterministic mine yields a byte-identical
restoration of the a3ac653 original. Caught by reading the artifact,
not assuming.
bench/control_ab.py + `make control-ab` (gated on judge-self-test
as a make dependency — instrument gate cannot be skipped). Same
model both arms; gold = target-article text by target_root; Arborist
[E…] scaffolding stripped (blinding — format can't betray the arm);
Arborist UNGROUNDED credited as honest abstention; hermetic Opus
judge; deterministic aggregate; self-auditing JSONL; threats-to-
validity printed in the report.
N=2 smoke: clean end-to-end, 0 JUDGE_ERROR — and already surfaced a
case AGAINST the treatment (solo correctly ABSTAINED; Arborist
HYBRID-WRONG). The instrument can falsify the Arborist value claim;
that is the point. n=2 proves nothing (report says so) — verdict
needs a real N.
fox ruled judge = Opus via `claude -p`. bench/judge.py:
hermetic (`env -u CLAUDECODE claude -p`, fresh process, context =
only (Q, answer, gold) — no arm label, no Arborist context, no
session), blinded-by-caller, reference-grounded against the fixed
gold (ignore parametric knowledge), structured via FINAL_VERDICT=
sentinel parsed LAST-match.
Instrument-before-experiment gate worked: first cut parsed
first-match over the model's chain-of-thought → 0/3 self-test. The
judge REASONED correctly; the parser was the defect (+ two bad test
fixtures, my error). Hardened (sentinel contract + fixed fixtures),
re-verified: `make judge-self-test` = 4/4 on known-verdict triples
via real claude -p. The make target is the precondition gate; no
control run trusts the judge until it passes.
Threat to validity recorded, not hidden: same model family judging;
mitigated (blind + no-stake + reference-grounded) not eliminated —
different-family SOTA cross-check is the only full removal.
Next: bench/control_ab.py + `make control-ab` (Hermes-solo vs
Arborist, gold=target-article text, blinded, judged) — NOT yet
built; no broken make target shipped for it.
fox: should the judge get the same docs, or the 8 retrieved/cited?
Neither. Per-arm retrieved docs (a) are undefined for Hermes-solo
(no retrieval → breaks A/B symmetry), (b) re-import circularity
(judging the treatment vs its own retrieval), (c) measure citation
self-consistency (Arborist's verifier already does that), not
truth. Judge's reference = the mined question's fixed ground-truth
target/gold, identical for both arms, independent of retrieval —
the ground-truth-carrying fixture supplies exactly this. Recorded a
distinct, non-headline faithfulness ablation (judge Arborist vs its
own context) so it is never conflated with the control delta.
Index row unchanged (skim surface; ticket body is the design log).
I over-applied a proof-path rule. "No LLM-as-judge" guards Arborist's
*verifier* (an LLM in the proof path = the circularity Arborist
kills; #000049's poor reliability is a proof-path-model constraint).
The control experiment's judge is external science — outside BOTH
arms, scoring post-hoc, touching neither system's internals — so it
cannot break the experiment. fox: use a strong SOTA model
(stronger than both systems under test), blinded + reference-
grounded + transcript-logged. This lifts the wrongly-imposed
atomic-fact-only limit (judge handles essay grounding; #000049's
exclusion was right for the proof path, wrong for external science);
deterministic gold-key demoted to a cheap atomic cross-check.
Still fox calls: fixture population, which SOTA judge (API
cost/egress), spend. Scaffold; no build past gate-1.
The review's strongest point (§4.1) + fox: every measurement so far
is WITHIN Arborist — no control arm. Folded into #000057 (not a new
ticket): control = Hermes-solo; judge external/deterministic/
paraphrase-tolerant via a mined gold-fact answer key (resolves
circularity + no-LLM-judge + paraphrase at once); headline metric =
fabrication-vs-honest-abstention delta, not raw accuracy (common-
knowledge fixture would falsely show null — parametric memory).
Smallest-proof reorder: v1 = single-shot Hermes-solo vs Arborist on
a fabrication-bait atomic-fact fixture; recursive-drift loop is v2,
gated on v1. Atomic-fact only; essay-grounding = #000049's unsolved
NLI, OUT. Still fox calls: fixture population, gold-key shapes,
spend. Scaffold; no build past gate-1.
