Repository navigation
Circle STARK reference collection: actual proofs, complete timings and machine records #133
Replies: 1 comment
|
Published and merged in PR #134, main The README headline tables and reference guide are live. All 18 merged-main CI jobs and the Pages deployment passed at that commit. The four release assets were downloaded anonymously and checked against their published sizes and SHA256 digests. A fresh verification from the archives completed in 219.253 seconds including compilation: all 23 distinct LLM proofs checked twice, all 36 LLM mutation cases rejected, and all 12 classifier proofs checked by fresh proof-only processes with native caller reconstruction. Evidence and receipts. These are separate component observations: the retained H200 LLM comparison, a fresh Mac default classifier run, and retained same-request composition. They do not establish a newly measured full GPU end-to-end latency. No new paid GPU job was run. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Circle STARK reference runs and proof downloads
This collection contains 23 TinyLlama proofs + three composed classifier
proofs, plus nine classifier control proofs. Every proof is a real serialized
proof. The manifest lists every file’s byte count and SHA-256. CPU/GPU logs,
caller statements, enrollment, source hashes, resource samples and job receipts
are included. This is a collection of explicitly named component runs, not one
uninterrupted fresh GPU-plus-classifier run.
Download artifacts
Runs and machines
llm-gpu-oluavcvse2e-circle-p0l734x_composed-circle-9bdff736The H200 prover reports 3,119,974,617,088 bytes of guest-visible system memory,
which is not the job allocation. Its provider profile and worker limit are
recorded separately. The declared worker address-space cap is 225,485,783,040 bytes; telemetry is
sampled. L1’s CPU model and cgroup CPU quota were not captured. No exclusive physical CPU host or hardware
attestation is claimed. Exact GPU image:
nvidia/cuda@sha256:e542739fcaa4f45da5add8c4cf5769783a61628b2518304a5bbe4ace468b8c8f.The separate Linux verification job uses HF
cpu-upgrade(8 provider vCPUs /32 GB). Raw logs preserve observed values; the provider profile is a declaration.
L1: complete TinyLlama response
The pinned
TinyLlama-1.1B-Chat-v1.0uses disclosedy-attnout-w12quantization.It proves “The capital of France is Paris.” from the fixed France question:
eight output tokens including EOS, 22 transformer layers plus final norm/head,
23 proofs, 22 residual links and seven generated-input links.
All 69 proof/statement/claim files are byte-identical between backends. The
matched ratios are 10.15× for the session timer and 4.36× for complete commands.
These are one CPU-then-GPU pair with shared caches, not latency distributions.
Generated constraints run on CUDA with zero constraint-evaluation fallback;
witness and interaction generation remain on CPU.
Do not add these rows: some are nested and others belong to separate jobs or
postflights. Lower-level FRI, Merkle and constraint timers are retained in each segment’s
result.point.phasesand in the numeric receipt. These profiler timers overlap;they are not an additive wall-time decomposition.
Native and Linux verifiers share SASH’s crypto implementation but
have different public preprocessing; their times are not a matched speedup.
Both require independently admitted model/shape/calibration roots and token case.
Exact statement, GPU method.
C1: default classifier workflow
C1 freshly trains and exports the frozen models, builds the locked prover and
proof-only verifier, then generates and verifies all nine control proofs.
Dependency caches are populated; the Rust target directory starts empty. This
is one local reference observation, not the repeated HF benchmark.
prove_secsPublic statement reconstruction, native reruns, host program construction,
18 cross-request checks, three roster checks and controller bookkeeping remain
inside the 93.50 s campaign; they were not individually timed. Internal stage
rows are nested inside the 50.03 s prover total; do not add both levels.
An all-pass verdict is separate from proof validity. The frozen MLP always
allows at this input limit; the CNN blocks the ordinary LLM prompt.
Each C1 proof
The raw artifact kind
encoderdenotes the contextual byte CNN.The repeated HF CPU comparison is a separate experiment:
eight workflows per backend on AMD EPYC 7R13 /
cpu-upgrade, with differentper-process costs and no training inside its sample timer. Its 298.88 s warm
Circle median must not be mixed with C1’s Mac stage observations.
J1: same-request composition
The retained-input default command takes 266.29 s. Its native LLM workflow
is 216.87 s, including a 115.53 s build, two fresh honest checks and all 36
LLM mutation cases. The classifier controller is 44.12 s; it generates three
new Circle proofs and runs binding checks. Three outer-join proof checks and
host hashing/binding complete the enclosing workflow. The nested stages and
root times are retained; no sum is presented as a fresh GPU end-to-end latency.
The three classifier proofs total 6,020,695 bytes; their complete prover
commands total 18.854 s, final STWO proving timers 9.083 s,
and inner proof-only verifier processes 0.421 s. The outer join verifies
those proofs again in 0.167 s of fresh verifier processes,
and rejects all 16 changed bindings/rosters.
All 23 LLM proofs and 22 links verify freshly. Verdicts
[1,1,0]give integritypass / policy release false. The LLM proof bytes equal L1’s published files;
the composition uses the exact fixed prompt, response and model. The classifiers
inspect the prompt; the response is bound into context, not classified. This
is host composition, with no recursive proof, hiding or hardware claim.
Download and verify
The artifact release is a research proof collection, not a qualified
software release. Its tag identifies the repository baseline before this
reference-documentation change; it is not every component’s measured commit. Use a checkout
containing this guide and
scripts/verify_reference.py; the Discussion pins theverification utility’s publication commit. No weights, witnesses, trusted remote
executable or HF credential are needed. Native verification requires the
Mac/Rust prerequisites, uv 0.9.18,
Python 3.12.12, the train extra for classifier public-input reconstruction,
and 12 GiB free plus build caches.
Obtain the manifest digest from this guide or the Discussion through a trusted
channel. A checksum delivered only by an untrusted prover is not enrollment.
The command first checks the externally pinned manifest, exact archive rosters
and every member hash. It rebuilds the source-pinned native verifier, checks the
LLM bundle twice with all 36 mutation cases, rebuilds the proof-only STWO
verifier and freshly checks all 12 classifier proofs, reconstructing their
expected public inputs. The composed request is checked against the fixed LLM
case. A hash-only success cannot produce the cryptographic success result.
Expected final counts: 23 distinct LLM proofs, two complete honest LLM
checks, nine classifier controls + three composed classifier proofs.
Failures return nonzero and preserve generated logs. The helper launches no
cloud job and downloads no model. Initial pinned Rust/SASH dependencies may need
network access; the proof workers themselves are sandboxed without networking.
Recompute the published numbers from the extracted raw records:
Source identities by component
The manifest’s
source_commitis the repository baseline for packaging thiscollection, not a statement that every historical component ran at that commit.
ef09216eabee044c24209f555df39e277ee6dfaa74bc2db33df5cf4d1e5a5b101b30fecba122960feb95bfa887299d1637550228d42086c27cbb2b3b; unchanged SIMD, inventoried generated CUDA patch9bea024dc7d4604f17d53c915b64bc4a23c408b5, with the new timing observer captured byte for byte before execution;timing/protocol.jsonpins every tracked pipeline filef613c611cefa9037320082b73777f26c9ec45990plus the archived per-file source snapshot incomposition/protocol.json; these are the default changes later merged as9bea0245fe56cdb459366850e83162769f8373451131b7d; crypto source identities are checked independently of the documentation commitBinary hashes, patches, model revisions, graph/preprocessing roots and concrete
parameters are retained with their own runs. Per-file snapshots identify local
changes; a bare commit label alone would not reproduce an uncommitted observer
or pre-merge controller. None of the historical measurement files is rewritten.
Concrete proof parameters
Both use Blake2s-based Merkle transcripts, but these are different relations,
implementations and concrete configurations. SASH reports a conjectured 100-bit
value; this is not independent security qualification. No equal-security speed
comparison follows. Each classifier enrollment also pins graph size, program
hash, preprocessing root and public-wire multiplicities.
Files, provenance and omissions
Manifest SHA-256:
2ea625c25d8305d6f006f6b492cffb066e93b5d7e8d92d7da6269156c5c1e6e8.The actual proof files and the transport archives have different sizes. The
classifier bundles contain hex-encoded JSON proof bytes; they are not compact
binary proofs. Rejection directories retain repeated copies of honest proofs
with changed public statements, so archive bytes are not a unique-proof sum.
The collection excludes credentials, private material, model checkpoints,
witness archives, compiled executables and caches. Selected raw logs are copied
without transformation; the manifest pins their original bytes. Source and
model revisions, licences and enrollment must be checked separately from hashes.
The LLM uses pinned upstream SASH source
eb95bfa887299d1637550228d42086c27cbb2b3b; SIMD is unchanged, CUDA has therecorded generated-kernel patch. The native LLM verifier identity is
8567c911d686b13b09032b75e1fc041e6bbcc9407cb2755b60746112102b342a.The classifier backend/compiler identity is
c035913b3196640797273bb0c22c2e69b09dee2a790d128be51287a6351ded7d.All weight/threshold/data identities remain frozen. The proof parameters differ
between SASH and direct STWO; this collection claims no equal-security ranking,
independent security review, production policy quality or hardware provenance.
The repository receipt
contains recomputed metrics, anonymous-download hashes, the fresh cryptographic
acceptance and the retained initial quota failure. The proof-verification utility
is pinned at commit
5fe56cdb459366850e83162769f8373451131b7d. The exact executedtiming observer is retained as
timing/recorder-used.py; its source hash matchesthe pre-run protocol. The maintained observer only removes its unused import.
Cryptographic acceptance of this exact archive set
Fresh source-based verification passes in 219.25 s, including native compilation and all 36 LLM mutations: 23 distinct LLM proofs checked twice, nine classifier controls and three composed classifier proofs checked by fresh proof-only processes. The proof archives were then downloaded without authentication; every byte count and SHA-256 matches. The first verification-helper attempt passed the LLM suite but charged input archives to its generated-work quota and stopped before classifier checks. That failure is retained; the pinned utility corrects the quota scope. No failed proof is labelled accepted. No new paid job was launched.
All reactions