feat: extend Jolt claim protocol formulas by markosg04 · Pull Request #1597 · a16z/jolt

markosg04 · 2026-06-09T19:20:35Z

Part of the draft Jolt prover stack.

Base: prover-stack/01-field-inline-tracing
Head: prover-stack/02-jolt-claims-protocol-formulas

Extends the Jolt claim formulas, including field-inline protocol formulas and supporting specs.

Validation before submission:

cargo fmt -q -- --check
cargo metadata --no-deps --format-version 1
local gh stack push simulation against a temporary bare remote
forbidden-path scan for handoffs/, old STACK.md, stack/, and old stack workflow

0xAndoroid · 2026-06-17T04:41:12Z

Stack maintenance — Pika is rebasing this PR onto #1596 (prover-stack/01-field-inline-tracing) and fixing CI as part of the prover stack (#1596–#1606). Rebase + CI summary to follow.

github-actions · 2026-06-17T05:08:56Z

Claude code review session started: https://claude.ai/code/session_01FNR82Crof9atPbqeKS2a11

0xAndoroid · 2026-06-17T05:09:01Z

Stack maintenance — rebased onto #1596 (prover-stack/01-field-inline-tracing); CI green. Marked ready for review (claude-review-request).

📚 Stack A: #1596 → #1597 (this PR) → #1598 → #1599 → #1600 → #1601 → #1602 → #1603
📚 Stack B: #1596 → #1597 (this PR) → #1604 → #1605 → #1606

#1596 and #1597 are shared by both stacks; each PR targets the one before it.

Conflict resolution

crates/jolt-claims/src/protocols/jolt/formulas/bytecode.rs — kept feat: add field-inline tracing support #1596's split address/cycle read-RAF flow and committed-program cycle output; layered feat: extend Jolt claim protocol formulas #1597's boolean-point / stage-value evaluation helpers on top; restored the address-phase virtual opening helper.
crates/jolt-claims/src/protocols/jolt/formulas/claim_reductions/advice.rs — kept feat: add field-inline tracing support #1596's shared PrecommittedClaimReduction scheduling as the source of truth; retained feat: extend Jolt claim protocol formulas #1597's explicit cycle-phase input-opening helper.
crates/jolt-claims/src/protocols/jolt/formulas/committed_openings.rs — kept feat: add field-inline tracing support #1596's final-opening order with committed-program chunks / program-image support; preserved feat: extend Jolt claim protocol formulas #1597's distinct proof commitment order.
crates/jolt-claims/src/protocols/jolt/formulas/error.rs — unioned feat: add field-inline tracing support #1596's precommitted/bytecode chunking errors with feat: extend Jolt claim protocol formulas #1597's evaluation-domain overflow error.

Follow-up CI fix (4344fe392c) — restored stacked enum variants/helpers required by #1596's committed-program formulas, restored the split booleanity helpers expected by verifier code, and added the small jolt-poly helpers used by #1597's formula evaluators.

Local validation

cargo fmt -q
cargo clippy --all --features host -q --all-targets -- -D warnings
cargo clippy --all --features host,zk -q --all-targets -- -D warnings
cargo build -p jolt-core -q
cargo nextest run -p jolt-core muldiv --cargo-quiet --features host
cargo nextest run -p jolt-core muldiv --cargo-quiet --features host,zk

cargo build --all --features host fails locally on macOS in zeroos-runtime-musl with a Mach-O section-specifier error. The targeted Jolt builds/checks above pass, and GitHub CI is green.

moodlezoup

Reviewed the Rust changes (specs/*.md skimmed as docs). This is verifier-side claim scaffolding for the prover stack; most new helpers are exercised only by their own tests so far. Opening-point orderings, the proof_commitment_order reorder (now matching the prover's all_committed_polynomials), and the SameEvaluationAs refactor all check out. A few notes below.

Minor / non-blocking:

jolt-poly/src/dense.rs: sumcheck_round_eval_with_order has no callers (all 25 sites use sumcheck_round_eval). Fine if a later stack entry needs it; otherwise drop it.
field_inline/formulas/dimensions.rs: read_write_opening_point / validate_phase_split / FieldRegistersReadWriteOpeningPoint are a near-verbatim copy of the jolt-side ReadWriteDimensions versions. Consider sharing the phase-split helper.
jolt-poly/src/eq.rs: boolean_index_msb's index.checked_shl(1)? guards the shift amount (always 1), not value overflow, so it never returns None on >64-var points. Not reachable from current callers, but the guard is misleading.

Generated by Claude Code

moodlezoup · 2026-06-17T05:21:04Z

+where
+    F: Field,
+{
+    Polynomial::new(EqPolynomial::<F>::evals(stage1_cycle_binding, None))


hamming_booleanity_eq_cycle_polynomial builds the eq table without reversing its cycle point, unlike every sibling (ra_virtualization_eq_cycle_polynomial here at L363, the instruction version, and eq_address_cycle_polynomial), which all .rev() first. The prover treats RAM hamming-booleanity and RAM ra-virtualization identically (both GruenSplitEqPolynomial::new(&params.r_cycle.r, LowToHigh) over a big-endian r_cycle), so the verifier helpers should be symmetric. The only test here is tautological (evals(x) == evals(x)) and there's no live caller yet. Please confirm the intended orientation of stage1_cycle_binding before this is wired up — if the caller passes the same big-endian point the other helpers receive, this table will be bit-reversed relative to the hamming_weight polynomial it multiplies.

Generated by Claude Code

moodlezoup · 2026-06-17T05:21:08Z

+    [
+        outer_opening(FieldInlineVirtualPolynomial::FieldRs1Value),
+        outer_opening(FieldInlineVirtualPolynomial::FieldRs2Value),
+        outer_opening(FieldInlineVirtualPolynomial::FieldRdValue),
+        outer_opening(FieldInlineVirtualPolynomial::FieldProduct),
+        outer_opening(FieldInlineVirtualPolynomial::FieldInvProduct),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::Add,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::Sub,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::Mul,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::Inv,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::AssertEq,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::LoadFromX,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::StoreToX,
+        )),
+        outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(
+            FieldInlineOpFlag::LoadImm,
+        )),
+    ]


This re-types the exact 13-entry sequence already declared in FIELD_INLINE_SPARTAN_OUTER_R1CS_INPUTS above, and the test just asserts the two are equal via map. Map the const directly:

Suggested change

[

outer_opening(FieldInlineVirtualPolynomial::FieldRs1Value),

outer_opening(FieldInlineVirtualPolynomial::FieldRs2Value),

outer_opening(FieldInlineVirtualPolynomial::FieldRdValue),

outer_opening(FieldInlineVirtualPolynomial::FieldProduct),

outer_opening(FieldInlineVirtualPolynomial::FieldInvProduct),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::Add,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::Sub,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::Mul,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::Inv,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::AssertEq,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::LoadFromX,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::StoreToX,

)),

outer_opening(FieldInlineVirtualPolynomial::FieldOpFlag(

FieldInlineOpFlag::LoadImm,

)),

]

FIELD_INLINE_SPARTAN_OUTER_R1CS_INPUTS.map(outer_opening)

Generated by Claude Code

moodlezoup · 2026-06-17T05:21:10Z

-    ///
-    /// Consumers must enforce these in the same constraint system as the input and output
-    /// claims; treating them as metadata can leave alternate opening chains unbound.
    pub consistency: Vec<JoltConsistencyClaim<F>>,


This drops the soundness-contract warning that previously documented consistency (consumers must enforce these in the same constraint system; treating them as metadata can leave alternate opening chains unbound). The field's behavior is unchanged, but that's a load-bearing, easy-to-misuse invariant — worth keeping the doc comment.

Generated by Claude Code

github-actions · 2026-06-17T16:07:30Z

Benchmark comparison (crates)

group                           main_run                               pr_run
-----                           --------                               ------
EqPolynomial::evals/17          1.09      2.1±0.03ms        ? ?/sec    1.00  1912.8±22.16µs        ? ?/sec
EqPolynomial::evals/19          1.09      7.9±0.06ms        ? ?/sec    1.00      7.3±0.08ms        ? ?/sec
EqPolynomial::evals/20          1.07     16.4±0.19ms        ? ?/sec    1.00     15.3±0.15ms        ? ?/sec
EqPolynomial::evals/22          1.09     62.9±0.44ms        ? ?/sec    1.00     57.5±0.28ms        ? ?/sec
EqPolynomial::evaluations/14    1.00   622.6±11.08µs        ? ?/sec    1.26   787.0±17.78µs        ? ?/sec
EqPolynomial::evaluations/18    1.10      7.2±0.10ms        ? ?/sec    1.00      6.5±0.07ms        ? ?/sec
EqPolynomial::evaluations/20    1.07     30.6±0.14ms        ? ?/sec    1.00     28.6±0.28ms        ? ?/sec
Fr::mul_u128                    1.16     25.0±0.11ns        ? ?/sec    1.00     21.5±0.17ns        ? ?/sec
Fr::mul_u64                     1.00     12.3±0.05ns        ? ?/sec    1.08     13.3±0.12ns        ? ?/sec
Fr::to_bytes                    1.13     20.1±0.32ns        ? ?/sec    1.00     17.7±0.22ns        ? ?/sec
Polynomial::bind/14             1.19    154.4±2.99µs        ? ?/sec    1.00    130.2±5.10µs        ? ?/sec
Polynomial::bind/18             1.11      2.2±0.03ms        ? ?/sec    1.00  1953.4±76.56µs        ? ?/sec
Polynomial::bind/20             1.10      8.5±0.05ms        ? ?/sec    1.00      7.7±0.07ms        ? ?/sec
Polynomial::evaluate/20         1.08     44.4±0.99ms        ? ?/sec    1.00     40.9±1.04ms        ? ?/sec
append_bytes/Blake2b/32B        1.00     49.1±2.22ns        ? ?/sec    1.10     53.9±4.45ns        ? ?/sec
append_bytes/Keccak/256B        1.10   914.2±19.32ns        ? ?/sec    1.00   832.5±20.28ns        ? ?/sec
append_bytes/Keccak/32B         1.00     42.9±0.46ns        ? ?/sec    1.13     48.3±3.87ns        ? ?/sec
append_bytes/Poseidon/32B       1.05     97.3±0.75µs        ? ?/sec    1.00     92.3±3.10µs        ? ?/sec
challenge/Blake2b               1.14   734.2±47.59ns        ? ?/sec    1.00   645.3±12.24ns        ? ?/sec
challenge/Keccak                1.11    439.6±7.61ns        ? ?/sec    1.00    395.9±5.94ns        ? ?/sec
challenge/Poseidon              1.07     48.8±0.37µs        ? ?/sec    1.00     45.6±0.60µs        ? ?/sec
g1_add                          1.12    464.3±2.22ns        ? ?/sec    1.00    414.7±5.31ns        ? ?/sec
g1_deserialize_bincode          1.05      9.7±0.09µs        ? ?/sec    1.00      9.2±0.15µs        ? ?/sec
g1_double                       1.10    233.6±1.72ns        ? ?/sec    1.00    212.0±1.93ns        ? ?/sec
g1_msm/1024                     1.10     14.0±0.03ms        ? ?/sec    1.00     12.8±0.28ms        ? ?/sec
g1_msm/16                       1.11   598.5±10.52µs        ? ?/sec    1.00   539.3±18.17µs        ? ?/sec
g1_msm/256                      1.11      4.7±0.01ms        ? ?/sec    1.00      4.3±0.01ms        ? ?/sec
g1_msm/4                        1.12    284.6±2.92µs        ? ?/sec    1.00    254.2±3.16µs        ? ?/sec
g1_scalar_mul                   1.11     72.2±0.90µs        ? ?/sec    1.00     65.1±1.40µs        ? ?/sec
g1_serialize_bincode            1.10    120.9±2.04ns        ? ?/sec    1.00    109.9±1.53ns        ? ?/sec
g2_msm/256                      1.12     17.5±0.22ms        ? ?/sec    1.00     15.6±0.06ms        ? ?/sec
g2_msm/4                        1.11    865.5±7.70µs        ? ?/sec    1.00    782.0±9.77µs        ? ?/sec
g2_msm/64                       1.12      6.3±0.02ms        ? ?/sec    1.00      5.6±0.01ms        ? ?/sec
g2_scalar_mul                   1.14   366.1±13.10µs        ? ?/sec    1.00    321.4±2.04µs        ? ?/sec
gt_scalar_mul                   1.00    777.8±9.26µs        ? ?/sec    1.15    892.0±6.27µs        ? ?/sec
multi_pairing/16                1.09      6.4±0.01ms        ? ?/sec    1.00      5.8±0.14ms        ? ?/sec
multi_pairing/2                 1.00  1071.9±22.49µs        ? ?/sec    1.17  1251.1±26.60µs        ? ?/sec
multi_pairing/4                 1.07  1972.6±18.05µs        ? ?/sec    1.00  1846.3±33.22µs        ? ?/sec
multi_pairing/8                 1.09      3.5±0.01ms        ? ?/sec    1.00      3.2±0.07ms        ? ?/sec
pairing                         1.00   810.7±42.29µs        ? ?/sec    1.14    923.8±8.00µs        ? ?/sec
pedersen_commit/1024            1.10     14.1±0.04ms        ? ?/sec    1.00     12.8±0.13ms        ? ?/sec
pedersen_commit/16              1.11    672.9±4.10µs        ? ?/sec    1.00    607.8±9.07µs        ? ?/sec
pedersen_commit/256             1.10      4.8±0.01ms        ? ?/sec    1.00      4.3±0.02ms        ? ?/sec
pedersen_commit/4               1.12    356.2±2.30µs        ? ?/sec    1.00    318.7±3.26µs        ? ?/sec

github-actions Bot added spec Tracking issue for a feature spec implementation PR contains implementation of a spec labels Jun 9, 2026

0xAndoroid force-pushed the prover-stack/01-field-inline-tracing branch 3 times, most recently from 18ed880 to 9fba45a Compare June 15, 2026 21:52

0xAndoroid mentioned this pull request Jun 17, 2026

feat: add field-inline tracing support #1596

Merged

0xAndoroid force-pushed the prover-stack/01-field-inline-tracing branch from 27f1951 to 8f437ff Compare June 17, 2026 04:29

0xAndoroid force-pushed the prover-stack/02-jolt-claims-protocol-formulas branch from 941c1b4 to 4344fe3 Compare June 17, 2026 04:55

0xAndoroid marked this pull request as ready for review June 17, 2026 05:08

0xAndoroid added the claude-review-request Request a review from Claude Code label Jun 17, 2026

This was referenced Jun 17, 2026

feat: add R1CS extension foundations #1598

Open

feat: add Jolt witness crate #1604

Merged

moodlezoup reviewed Jun 17, 2026

View reviewed changes

0xAndoroid mentioned this pull request Jun 17, 2026

docs: add akita lattice specs #1606

Open

Base automatically changed from prover-stack/01-field-inline-tracing to main June 17, 2026 15:53

markosg04 and others added 4 commits June 17, 2026 15:18

feat: extend Jolt claim protocol formulas

dc31a0b

fix: preserve stacked claim formula APIs

ce66ad5

ci: cover field-inline verifier paths

c16f229

ci: run field-inline tracer tests after CLI install

62aab06

0xAndoroid force-pushed the prover-stack/02-jolt-claims-protocol-formulas branch from 91dbcf2 to 62aab06 Compare June 17, 2026 19:18

moodlezoup merged commit caa7a07 into main Jun 23, 2026
28 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: extend Jolt claim protocol formulas#1597

feat: extend Jolt claim protocol formulas#1597
moodlezoup merged 4 commits into
mainfrom
prover-stack/02-jolt-claims-protocol-formulas

markosg04 commented Jun 9, 2026

Uh oh!

0xAndoroid commented Jun 17, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jun 17, 2026

Uh oh!

0xAndoroid commented Jun 17, 2026 •

edited

Loading

Uh oh!

moodlezoup left a comment

Uh oh!

moodlezoup Jun 17, 2026

Uh oh!

moodlezoup Jun 17, 2026

Uh oh!

moodlezoup Jun 17, 2026

Uh oh!

github-actions Bot commented Jun 17, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

markosg04 commented Jun 9, 2026

Uh oh!

0xAndoroid commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jun 17, 2026

Uh oh!

0xAndoroid commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

moodlezoup left a comment

Choose a reason for hiding this comment

Uh oh!

moodlezoup Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

moodlezoup Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

moodlezoup Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark comparison (crates)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

0xAndoroid commented Jun 17, 2026 •

edited

Loading

0xAndoroid commented Jun 17, 2026 •

edited

Loading

github-actions Bot commented Jun 17, 2026 •

edited

Loading