Collaborative Intelligence

State of knowledge

Communication protocols between agents, evidence exchange versus conclusion sharing, correlated errors, and how disagreement should change a conclusion.

This track draws on 16 published rounds, 31 hypotheses and 46 experiments. 101 reported findings come from published rounds and have not been independently replicated.

Reported findings

• F54.S1 (deterministic, E54-A): Over all 27 states of gold and two verdicts, agreement routing is unsafe in 4 states, exactly when both reads share the same wrong decisive verdict, flag only review is unsafe in 4 states, and rewriting review is unsafe in 12 states including 6 new ones. The routing identity holds on 1,672 of 1,672 recorded pairs. (Round 54)
• F54.S2 (deterministic, E54-A): All 120 track H certificates and all 32 SEC windows check; all 32 SEC accessions and filers are new against the recorded prior set through Round 53. (Round 54)
• F54.E1 (live, E54-B): Witness D released 35 unsafe verdicts on 152 items (Wilson interval 0.1705 to 0.3033), including a wrong FAIL on 22 of 24 held out near miss items and on 12 of 24 partial period certificates, all 35 on track H. (Round 54)
• F54.E2 (live, E54-B): Routing Witness D with its own resample left 29 unsafe verdicts; routing with Witness B left 3; with Witness C, 8; with Witness D's deliberate read, 14. Spearman rho between disagreement and unsafe verdicts removed was 0.9495. (Round 54)
• F54.E3 (live, E54-B): Removing the HOLD case from the deliberate prompt raised unsafe verdicts from 15 to 26 (p = 0.003418). (Round 54)
• F54.E4 (live, E54-C): A seeded wrong proposal was not kept more often than the same reviewer's own blind error for any of the three reviewers. (Round 54)
• F54.E5 (live, E54-C): Flag only review created 0 new unsafe releases and left 5 unsafe verdicts; rewriting review created 0 new unsafe releases and left 11. (Round 54)
• F54.E6 (live, E54-D): On Witness D, a stored conclusion produced 33 unsafe verdicts against 35 (p = 0.753906), and a stored procedure produced 26 against 35 (p = 0.022461); the ablation was 28 against 26 (p = 0.753906). On Witness C the same contrasts were 17 against 9 and 0 against 9. (Round 54)
• F54.E7 (live, E54-B): 393 of 395 unsafe decisive quotes copied the packet exactly (0.9949, interval 0.9817 to 0.9986). (Round 54)
• F54.E8 (live, E54-B): The C1 pattern held in all 6 domain groups and on all 3 surfaces. Temperature 0 repeats agreed on 134 of 152, 146 of 152 and 136 of 152 items for Witnesses B, C and D. (Round 54)
• F53.S1 (deterministic, E53-A): Over 27 states of gold and two verdicts, agreement routing is unsafe in 4 states, exactly when both reads share the same wrong decisive verdict, flag only review is unsafe in 4 states, and rewriting review is unsafe in 12 states including 6 new ones. The routing identity holds on 2067 of 2067 recorded pairs. (Round 53)
• F53.S2 (deterministic, E53-A): All 36 track R prompts are identical to Round 49 and Round 52; all 120 track G certificates and all 32 SEC windows check. (Round 53)
• F53.E1 (live, E53-B): Witness D released 38 unsafe verdicts on 188 items (Wilson interval 0.1510 to 0.2652), including a wrong FAIL on 9 of 9 Round 49 near miss items and on 24 of 24 held out near miss items, and 29 unsafe verdicts on track G. (Round 53)
• F53.E2 (live, E53-B): Routing Witness D with its own resample left 35 unsafe verdicts; routing with Witness B left 4; with Witness C, 18; with Witness D's deliberate read, 13. Spearman rho between disagreement and unsafe verdicts removed was 0.968. (Round 53)
• F53.E3 (live, E53-B): Removing the HOLD case from the deliberate prompt raised unsafe verdicts from 13 to 32 on 187 paired items. (Round 53)
• F53.E4 (live, E53-C): A seeded wrong proposal was not kept more often than the same reviewer's own blind error for two of three reviewers. (Round 53)
• F53.E5 (live, E53-C): Flag only review created 0 new unsafe releases and left 12 unsafe verdicts; rewriting review created 0 new unsafe releases and left 21. (Round 53)
• F53.E6 (live, E53-D): On Witness D, a stored conclusion produced 44 unsafe verdicts against 38, and a stored procedure produced 33 against 38, both above the preregistered p threshold. On Witness C the same contrasts were 27 against 18 and 2 against 18. (Round 53)
• F53.E7 (live, E53-B): 526 of 529 unsafe decisive quotes copied the packet exactly (0.9943, interval 0.9835 to 0.9981). (Round 53)
• F52.S1 (deterministic, E52-A): Over 81 states of gold and three verdicts, unanimity, recovery and majority are unsafe in 4, 20 and 28 states, strictly nested, and every unsafe recovery state needs two reads sharing the same wrong decisive verdict. (Round 52)
• F52.S2 (deterministic, E52-A): All 36 track R prompts are identical to Round 49; all 16 table claims and their exclusion traps and two world completions recompute; all 24 SEC windows contain their definition sentence. (Round 52)
• F52.E1 (live, E52-B): Witness D reproduced its Round 49 verdicts on 36 of 36 items, including a wrong FAIL on 9 of 9 near miss items. (Round 52)
• F52.E2 (live, E52-B and E52-C): Two Witness D resamples and a Witness D self validation gave the same wrong FAIL on 9 of 9 near miss items; self validation changed none of 75 reads. (Round 52)
• F52.E3 (live, E52-B): At two calls, routing Witnesses C and D released 6 unsafe verdicts, routing Witness C with its own resample 5, and routing Witness D with its own resample 9. (Round 52)
• F52.E4 (live, E52-B): One Witness D call with room to argue each verdict released 2 unsafe verdicts on 69 usable items, moving 5 near miss FAILs to HOLD; seven of its outputs were empty. (Round 52)
• F52.E5 (live, E52-C): Witness C as validator moved 2 of 7 wrong Witness D reads to HOLD and turned one correct HOLD into PASS; recovery released 5 unsafe verdicts against 0 for three reader unanimity. (Round 52)
• F52.E6 (live): Witness B alone released no unsafe verdict, with 2 false holds; no reader made a decisive error on the 16 table claims or the 24 SEC claims, and Witness A's blind audit agreed with all 40 of their labels. (Round 52)
• F51.S1 (deterministic, E51-A): An exact evidence span check admits all 64 scripted wrong status mutations of the authored code and table ledgers. (Round 51)
• F51.S2 (deterministic, E51-A): Running each implementation on the specification's own examples admits all four subtle defects as PASS and routes all four obvious defects. (Round 51)
• F51.S3 (deterministic, E51-A): Across 81 states, preservation loses 12 correct decisive repair states; independent recovery restores four and adds eight unsafe states, every one with the rewrite and the independent reader sharing the same wrong decisive verdict. (Round 51)
• F51.E1 (live, E51-B): On 32 authored code and table items, Witnesses B, C and D made no decisive error against execution or recomputation, catching all subtle defects and table traps. (Round 51)
• F51.E2 (live, E51-B): Witness B identified a wrong authored HOLD label (X01H) with a counterexample confirmed by a deterministic check; the blind label audit agreed with all 54 fixed labels, including the wrong one. (Round 51)
• F51.E3 (live, E51-B): Witness C read four of six SEC compliance claims as FAIL against fixed HOLD labels, wording the blind audit had flagged as admitting FAIL. (Round 51)
• F51.E4 (live, E51-C): The validator changed none of 53 natural baselines and kept HOLD on four items where its own blind read was FAIL. (Round 51)
• F51.E5 (live): One reader output and the first synthesis call were empty; neither was scored or repeated as an experimental read. (Round 51)
• F50.S1 (deterministic, E50-A): All 48 wrong status mutations retain valid quotes and pass the quote gate; the status preservation gate rejects all 48 and admits all 24 unchanged correct ledgers. (Round 50)
• F50.S2 (deterministic, E50-A): Across all 27 truth and paired verdict states, agreement routing leaves four unsafe states. (Round 50)
• F50.E1 (live, E50-B): Blind review agrees with 30 of 30 labels but flags named instrument versus generic approval wording on positive and cure variants. (Round 50)
• F50.E2 (live, E50-B): Decisive unsafe reads and same wrong decisive agreements are zero on all usable reads and pairs; joint false holds remain (five without procedure memory, four with it). (Round 50)
• F50.E3 (live, E50-B): On the 21 common items, paired review is correct on 16 with five false holds under both memory arms. (Round 50)
• F50.E4 (live, E50-C): Seven of 29 usable correct input rewrites become wrong HOLD statuses with exact quotes; the preservation gate routes all seven but does not improve the HOLD endpoint. (Round 50)
• F50.E5 (live, E50-C): From wrong inputs the validator repairs 22 of 30 ledgers; the preservation gate routes all 22, admits eight wrong unchanged statuses, and reduces final correctness from 22 to 12 of 30. (Round 50)
• F50.E6 (live): Five individual outputs are unusable (two empty, three truncated); none was repeated or scored as a HOLD. (Round 50)
• F49.1 (deterministic, E49-A): A quote check alone leaves 9 of 36 stale conclusions intact; adding a coverage check withholds all 36 and leaves all 36 correct notes intact. (Round 49)
• F49.2 (live, E49-B and E49-D): Witness D never adopts a stale stored conclusion (0 of 36); accuracy is 26 of 36 with no memory and 27 of 36 with a correct, stale, or guarded note. (Round 49)
• F49.3 (live): Witness D answers FAIL on 9 of 9 near miss items under every answer memory arm, including the correct note that stores HOLD with its reason. (Round 49)
• F49.4 (live): A procedure card raises Witness D to 30 of 36, with near miss items 5 of 9, unsafe verdicts 4 against 9, and false holds 2 against 1, in both domains. (Round 49)
• F49.5 (live): Witness C scores 28, 27, and 29 of 36 under no memory, stale, and guarded notes, with 0 stale follows. (Round 49)
• F49.6 (live): A wrong brief changes 0 of 36 provisional verdicts; a right brief changes 1, wrongly, and the lock blocks it. (Round 49)
• F49.7 (live, E49-C): Live note following is 0 of 24 across three note styles; five of ten CERTIFY exhibits rewritten to HOLD cite verified condition precedent text. (Round 49)
• F48.1 (deterministic, E48-A): R48 releases all five Domain A and all four Domain B gold cures that R47 holds, with zero unsafe releases under scripted M2. (Round 48)
• F48.3 (live, E48-B): On Witness C live Domain A ledgers, R47 and R48 are identical at 14 of 20 with 0 unsafe. On all 20 Witness B ledgers (15 completed post hoc), R48 improves correct open outcomes from 12 to 15 without an unsafe release. (Round 48)
• F48.4 (live, E48-B): Extract first provisional matches packet only at 18 of 20. V4 final is 14 of 20 against V3 at 12 of 20. Wrong resolver holds followed: V1 0, V3 4, V4 final 2. (Round 48)
• F48.5 (live, E48-D): On Domain B clinical packets, V1 is 16 of 16 and V4 final is 14 of 16. (Round 48)
• F48.6 (live, E48-B): All routing compositions escape zero unsafe outcomes. C1 with R48 and C1 with R47 both escape zero. (Round 48)
• F48.7 (live, E48-C): Live note following is zero on 24 exhibits. Ten CERTIFY exhibits become false holds. (Round 48)
• F47.1 (deterministic, E47-A): R47 releases zero of five gold cures held by R46 and retains the universal test against all ten scripted schedule threshold releases. (Round 47)
• F47.2 (deterministic, E47-A): Exploratory R47c releases all five gold cures but permits ten unsafe schedule threshold releases. (Round 47)
• F47.3 (live, E47-B): On Witness C ledgers, R47 improves correct open outcomes from 10 to 13 without an unsafe release; on all 20 Witness B ledgers (6 completed post hoc), R47 improves from 12 to 13, also without an unsafe release. (Round 47)
• F47.4 (live, E47-B): All routing compositions escape zero unsafe outcomes. C1 with R47 routes 17 of 20, compared with 18 of 20 for C1 with R46. (Round 47)
• F47.5 (live, E47-B): Independent Witness D is less accurate with packet and brief than packet alone, 11 against 17 correct, and follows wrong resolver holds as often with both as with brief alone. (Round 47)
• F47.6 (live, E47-C): Live note following is zero on 24 exhibits. Seven CERTIFY exhibits become false holds and the gate accepts all seven. (Round 47)
• F46.1 (deterministic, E46-A): On gold ledgers, R46 holds all 5 correctly read cures under the open resolver and turns all 5 into wrong FAILs under the closed resolver; R45, R45+F1, and R46c reproduce all 20 labels. (Round 46)
• F46.2 (deterministic, E46-A): Under scripted split condition mistakes, F1 removes the near miss counting release (M1, 10 to 0) and only R46 removes the schedule threshold release (M2, 10 to 0), at the price of holding every correct cure reading (M3, 5 of 5). (Round 46)
• F46.3 (deterministic, E46-A): A disagreement router releases an unsafe outcome only when both components share the error. (Round 46)
• F46.4 (live, E46-B): On Witness C's live ledgers, R45 was correct on 20 of 20, R46 on 15 of 20 (all 5 cures held, 0 unsafe under the open resolver), and R46c and R45+F1 on 19 of 20. Witness C never split a requirement, so no split condition release occurred on its ledgers. On Witness B's 20 ledgers (12 completed post hoc the same day), R45 made 3 split condition releases (wrong PASS) under the open resolver and R45+F1, R46, and R46c made none; R46 was correct on 10 of 20 against 12 for R45. (Round 46)
• F46.5 (live, E46-B): Both direct readers were correct on 17 of 20; all 5 of their unsafe verdicts were FAIL on near miss items. No direct reading or reviewer verdict was a wrong PASS. (Round 46)
• F46.6 (live, E46-B): Routing against the resolver escaped 0 unsafe outcomes at loads from 0.30 (R45) to 0.55 (R46); routing two readers against each other escaped 2 at a load of 0.20. (Round 46)
• F46.7 (live, E46-B): The reviewer was correct on 16 of 20 with the packet only (4 unsafe), 14 of 20 with the brief only (0 unsafe, 6 holds on PASS or FAIL items), and 20 of 20 with both. With the brief only, it followed a wrong resolver hold on 3 of 5 items; with both, on 0 of 5. (Round 46)
• F46.8 (live, E46-C): On 24 real exhibits, the validator followed a note quoting real window text on 0 of 24 and never identified the note as the error. (Round 46)
• F45.1 (deterministic, E45-A): on correct gold ledgers, the Round 44 rules are wrong on all 6 K2 items under both resolvers and release an unsafe FAIL on all 6 TS items under the open resolver; R45 reproduces all 60 labels. (Round 45)
• F45.2 (deterministic, E45-A): each scripted reader mistake defeats a different set of rules; a record scoped negative quoted as an exclusion and marked as covering defeats all four pre registered rule sets and is caught only by the locality test. (Round 45)
• F45.3 (deterministic, E45-A): on 24 SEC exhibits, a rewrite that follows the misleading note releases on all 12 HOLD exhibits and holds on all 12 CERTIFY exhibits; the unchanged gate routes all 12 releases to review and accepts all 12 holds. (Round 45)
• F45.4 (live, E45-B): the blind review agreed with 60 of 60 labels and raised 7 defect notes, one or more in every family, so every family is contested under the fixed rule. (Round 45)
• F45.5 (live, E45-B): on 60 ledgers, R45 with the open resolver was correct on 52 with 4 unsafe releases, against 36 correct and 11 unsafe under the Round 44 rules on the same ledgers; Witness C was correct on 30 of 30. (Round 45)
• F45.6 (live, E45-B): on the same ledgers, R44 was wrong and R45 right on 19 (9 credited record sets, 10 unscoped negatives no longer read as FAIL), and R44 right and R45 wrong on 3, all Witness B. (Round 45)
• F45.7 (live, E45-B): under R45, TC resolved to FAIL on 8 of 12 ledgers and TX, TS, and T on 0 of 12 each. (Round 45)
• F45.8 (live, E45-B): record sets without the scope test raised unsafe releases from 11 to 19; the scope test without record sets removed all unsafe releases but held 10 of 12 K2 ledgers. (Round 45)
• F45.9 (live, E45-B): the exploratory locality rule changed no unsafe outcome and lost 2 TC ledgers where Witness B marked the completeness record itself FAILS. (Round 45)
• F45.10 (live, E45-B): direct readers were correct on 46 of 60 with 14 unsafe FAILs, all on near miss items; the resolver was wrong and the reader right on 7 ledgers, the resolver right and the reader wrong on 13, and both wrong on 1. (Round 45)
• F45.11 (live, E45-B): Witness D, as a separate panel, was correct on 22 of 30 under R45 with 0 unsafe releases and 8 false holds, including all 6 TC items. (Round 45)
• F45.12 (live, E45-C): Witness C's 72 exhibit validator calls produced 72 correct outcomes; no rewrite followed the misleading note (0 of 24) and the gate routed 12 correct repairs to review. (Round 45)
• F45.13 (completion): all phases after the first 57 record set calls were completed in a same day, post hoc rerun with unchanged prompts and settings. (Round 45)
• F44.1 (deterministic, E44-A): all eight certificates pass; exactly the 8 T items are HOLD under an open record by two completion certificate, and under the record level resolver no T ledger can release FAIL without a verified exclusion quote, while the condition ledger releases FAIL from the near miss record alone. (Round 44)
• F44.2 (live, E44-B): the blind review agreed with 48 of 48 authored labels and reported packet defects in 6 of 8 families; 5 of 11 defect notes concern K packets, and the reviewer's own verdicts matched every T label. (Round 44)
• F44.3 (live, E44-B): under the open resolver, record level ledgers held 23 of 24 T readings and released an unsafe FAIL on 1 (Witness B, N03T, a record scoped negative quoted as an exclusion); on the same items the condition ledger released FAIL on 12 of 16 and declared open direct readers on 8 of 16. (Round 44)
• F44.4 (live, E44-B): record level ledgers resolved all 24 TC items to FAIL under both resolvers and all 24 T items to FAIL under the closed resolver; direct readers returned FAIL on all 16 TC readings. (Round 44)
• F44.5 (live, E44-B): on K, Witness C was correct on 8 of 8 under both resolvers; Witnesses B and D held 4 and 3 of 8 under the open resolver and released 2 and 3 unsafe FAILs under the closed resolver, mostly on requirements met only by two records together. (Round 44)
• F44.6 (live, E44-B): under the record level ledger no two witnesses shared an unsafe release on T; under the condition ledger and direct reading, Witnesses B and C erred on the same 4 of 8 T items. (Round 44)
• F44.7 (live, E44-C): across 144 validator calls on 24 SEC exhibits, no arm made an unsafe release; flagging kept all 24 spurious condition seeds as false holds, rewriting repaired all 48 seeded ledgers without changing any gold outcome, and the gate routed the 24 hold to release repairs to review. (Round 44)
• F43.1 (deterministic, E43-A): all seven certificates pass; a flag only validator cannot create an unsafe release, while a span valid rewrite turns each of the 12 gold HOLD exhibit ledgers into an unsafe release. (Round 43)
• F43.2 (live, E43-B): undeclared, direct readers released FAIL on 33 of 36 U and T decisions; declared open, U errors fell from 15 of 18 to 0 of 18 and T errors from 18 of 18 to 6 of 18, all FAIL on near miss items (S03T, S06T, S02T). (Round 43)
• F43.3 (live, E43-B): declared closed, direct readers were correct on 36 of 36 U and T decisions and 18 of 18 K decisions; K was also 18 of 18 under the open declaration. (Round 43)
• F43.4 (live, E43-B): extraction readers under the open resolver held 18 of 18 U items and released FAIL on 14 of 18 T items, marking the requirement UNMET from the near miss record; under the closed resolver they were correct on 35 of 36 with no unsafe release. (Round 43)
• F43.5 (live, E43-B): under the open convention, 5 of 6 direct unsafe releases coincided with the same witness's extraction unsafe release (independence 2.33); error correlation was positive for every role pair and every cross witness pair. (Round 43)
• F43.6 (live, E43-C): on 48 exhibit ledgers, the flag only arm made 0 unsafe releases and 12 false holds and repaired 12; the rewriting arm made 0 unsafe releases and 0 false holds and repaired 24; neither corrupted a correct ledger. (Round 43)
• Constructed example: a one-bit message that answers the release question perfectly leaves a 3/4 accuracy ceiling on a later test-status question under a stipulated uniform distribution. A possibility result, not model performance. (Round 17)
• Exact enumeration: 5, 2 and 1 of 15 partitions preserve every answer as the question family grows, needing at least 2, 3 and 4 message states. None of 60 deterministic recompressions restores a lost distinction. (Round 17)
• Behavioral influence does not establish a useful correction. (Round 15)
• Message value depends on the receiver's existing information and on other messages. (Round 15)
• A local communication benefit must still survive a complete-system comparison under measured resources. (Round 15)

Rounds in this track

1. Round 54 · 2026-11-11
   The Same Statements, New Cases: Held Out Replication of Domain General Propositions on New Packets and New Surface Forms

   Notebook
2. Round 53 · 2026-11-10
   Stated Plainly, Tested to Break: Domain General Candidate Propositions on Disagreement, Review, Memory and Evidence, on Held Out Mixed Cases

   When candidate propositions about disagreement, review, memory and evidence are stated in domain general language, which of them survive preregistered attempts to break them on leftover mixed cases that are not only contracts?

   Notebook
3. Round 52 · 2026-11-09
   The Second Reader or the Second Call? Matched Call Baselines for Agreement Routing, Validation and Recovery

   When a second or third model call improves a verdict, is it because a different reader looked, or only because another call was made, once every composition is compared with the same reader resampled, the same reader validating itself, and one call with room to argue each verdict, at the same number of calls?

   Notebook
4. Round 51 · 2026-11-08
   Valid Evidence, Wrong Verdict: Transfer Tests on Code Review and Raw Table Claims, with Independent Recovery

   Do the evidence, agreement and rewrite checks of the contract rounds transfer to a clearly different non-document task, code review against a specification and numerical claims against raw tables, and does routing a changed status to an independent reader recover correct repairs that a preservation gate would lose?

   Notebook
5. Round 50 · 2026-11-07
   When Both Readers Are Wrong: Open Removal Tests for Review and Rewriting

   Open removal tests, not another contract refinement: joint error when both roles can fail in the same direction, and rewrite gate behavior on genuinely wrong rewrites.

   Notebook
6. Round 49 · 2026-11-07
   Remember the Answer, or Remember the Procedure?

   When an independent reviewer reuses working memory, does a stale stored conclusion poison later decisions, does a deterministic revalidation guard remove the stale part, and does a reusable procedure help more than a stored answer?

   Notebook
7. Round 48 · 2026-11-06
   Extract First, or Brief First?

   Does a condition level discharge field release correctly read cures without weakening schedule threshold protection, and does extract first independent review reduce brief anchoring?

   Notebook
8. Round 47 · 2026-11-05
   When a Replacement Is Not a Cure: Supersession, Schedule Thresholds, and Independent Review

   Does explicit supersession release correctly read cures without weakening schedule threshold protection, and what must an independent reviewer see?

   Notebook
9. Round 46 · 2026-11-04
   Who Must Look Again? Disagreement Routing, False Hold Load, and What a Reviewer Must See

   When a resolver and a reader disagree and the item goes to review, what does a consistency test hold back, how much is sent to review and what escapes, and what must the reviewer see?

   Notebook
10. Round 45 · 2026-11-03
    The Rule or the Reader? Record Sets, Scoped Exclusions, and Where a Fixed Resolver Is Wrong and a Witness Is Right

    When the reader gets the records right, where is a fixed resolver rule wrong, and when the rule is changed with record sets and a scope test to fix that, what new mistakes does it let through?

    Notebook
11. Round 44 · 2026-11-02
    This Record Fails, or No Record Exists? Record Level Status Ledgers, Completeness Records, and Independently Reviewed HOLD Labels

    When an evidence ledger records what each record shows and asks separately whether anything excludes other records, do readers stop treating a defective record as proof that no valid record exists, and do the authored HOLD labels survive a blind independent review?

    Notebook
12. Round 43 · 2026-11-01
    Open Record or Closed Record? Declared Record Conventions, Near Miss Records Read as Negation, and Flag Only Against Rewriting Validators

    When the record convention is declared, do readers still read a missing or defective record as proof that an event did not happen, and should a validator flag ledger defects or rewrite them?

    Notebook
13. Round 17 · 8 October 2026
    When the Question Changes: Recoverable Evidence Under a Context Budget

    When a follow-up question changes which facts are decisive, does a compressed handoff that records dependencies and a declared scope improve supported completion over a strong contextualized prose summary, when both can reopen the same sources and all work is charged to the same budget?

    Notebook
14. Round 15 · 6 October 2026
    What makes a message worth sharing?

    Can a bounded mechanism for retaining and combining unresolved evidence improve Noetis after its additional costs and revision damage are measured?

    Notebook
15. Round 10 · October 2026
    Evidence exchange, selective revision, and a testable path

    What communication and revision protocol makes a correction trustworthy enough to reuse?

    Notebook
16. Round 3 · August 2026
    Hippocampal ripples and compositional inference

    Is replay a useful hypothesis for recombining evidence into new inferences?

    Notebook

Hypotheses

• H54.3Each of route(D, B), route(D, C) and route(D, D deliberate) has fewer unsafe outcomes than route(D, D resample 1), with at least two McNemar tests below 0.05.Supported
• H54.4Across eleven second reads paired with Witness D, Spearman correlation between disagreement (without gold) and unsafe outcomes removed by routing is at least 0.7.Supported
• H54.6For route(D, B), route(D, C) and route(D, D deliberate), unsafe outcomes removed relative to Witness D's read are at least the false holds added.Supported
• H54.12In at least 5 of 6 track H domain groups, route(D, B) and route(D, D deliberate) each have no more unsafe outcomes than route(D, D resample 1).Supported
• H54.13On at least two of the three surfaces (table, log, letter), route(D, B) and route(D, D deliberate) each have no more unsafe outcomes than route(D, D resample 1).Supported
• H53.2Replication: Witness D is FAIL on at least 7 of 9 Round 49 near miss items; its resamples reproduce at least 80 percent of those wrong decisive verdicts; its deliberate read moves at least 3 of the 9 to HOLD.Supported
• H53.3Each of route(D, B), route(D, C) and route(D, D deliberate) has fewer unsafe outcomes than route(D, D resample 1); at least two McNemar tests give p below 0.05.Supported
• H53.4Across eleven second reads paired with Witness D, Spearman correlation between disagreement (without gold) and unsafe outcomes removed by routing is at least 0.7.Supported
• H53.5route(D, D deliberate) beats route(D, D resample 1); removing the HOLD case from the deliberate prompt removes the benefit.Supported
• H53.6For route(D, B), route(D, C) and route(D, D deliberate), unsafe outcomes removed relative to Witness D's read are at least the false holds added.Supported
• H53.10At least 90 percent of unsafe decisive outputs with a nonempty quote copy the packet exactly, and the Wilson lower bound is at least 0.8.Supported
• H53.11Witness D releases at least 10 unsafe verdicts on track G; each reader's temperature 0 repeat agrees on at least 95 percent of usable items.Supported
• H53.12In at least 5 of 6 track G domain groups, route(D, B) and route(D, D deliberate) each have no more unsafe outcomes than route(D, D resample 1).Supported
• H52.3At two calls, routing Witnesses C and D has fewer unsafe outcomes than routing either reader with its own resample, pooled over all items.Rejected
• H52.4Routing Witnesses C and D does not dominate routing the better of the two readers with its own resample: it pays at least as many false holds.Supported
• H52.7Cross validation changes more of Witness D's wrong decisive reads to gold or HOLD than self validation.Supported
• H51.3In each non-document domain at least one reader returns a decisive verdict on an open dependency HOLD item.Rejected
• H51.4Agreement routing of Witnesses C and D yields no more unsafe outcomes than the better single reader in each domain, with a false hold cost.Inconclusive
• H50.1At least one same wrong decisive agreement survives routing in each domain under no memory.Rejected
• H50.5Removing the second read does not improve safety, though it can improve endpoint accuracy by avoiding false holds.Inconclusive
• H49.1A stale stored note lowers Witness D accuracy relative to no memory, and Witness D follows the stale conclusion on at least 9 of 36 items.Rejected
• H49.2A correct stored note is at least as accurate as no memory for Witness D.Supported
• H49.6Under a stale note, Witness C gives the stale wrong verdict on more items than under no memory.Rejected
• H48.5C1 with R48 escapes no more unsafe releases than C1 with R47.Supported
• H47.4C1 with R47 escapes no more unsafe releases than C1 with R46.Supported
• H46.3C1 escapes no more unsafe outcomes than the better of its two components alone, at a review load below 50 percent.Rejected
• H46.4Cross witness routing (C1) escapes no more unsafe outcomes than same witness routing (C2).Supported
• H45.6Where the resolver is wrong, the same witness reading directly is right more often than wrong.Supported
• H44.4Under the record level ledger, unsafe releases on T items do not coincide across witnesses.Inconclusive
• H43.3Once extraction readers can err unsafely, a witness's unsafe releases coincide across its two roles.Supported
• H10Exchanging evidence makes corrections more reusable than exchanging conclusions.Proposed

Experiments

• E54-AStructural diagnostic for routing, flag only review and rewriting review, with gold certificatesComplete
• E54-BIndependent, repeated, resampled and deliberate reads on new packets and new surface formsComplete
• E54-CRewriting review, flag only review and seeded wrong proposalsComplete
• E54-DStored conclusions, a domain general stored procedure, and an ablation of its open record stepsComplete
• E53-AStructural diagnostic for routing, flag only review and rewriting review, with replay and gold certificatesComplete
• E53-BBlind label audit and independent, repeated, resampled and deliberate readsComplete
• E53-CRewriting review, flag only review and seeded wrong proposalsComplete
• E53-DStored conclusion and stored procedure memory armsComplete
• E52-AStructural diagnostic for unanimity, majority, routing and recovery, with replay and gold certificatesComplete
• E52-BIndependent, resampled, deliberate and procedure card reads at matched call countsComplete
• E52-CCross validation, self validation and recovery against matched call baselinesComplete
• E51-AStructural diagnostic for evidence span, example execution, preservation and recovery gatesComplete
• E51-BIndependent reads on code review, raw table claims and SEC covenant claimsComplete
• E51-CValidation of natural baselines with replayed preservation and recoveryComplete
• E50-AStructural diagnostic for quote and status preservation gates and agreement routingComplete
• E50-BSecond read and procedure memory removal on contract and non-contract policy packetsComplete
• E50-CLive ledger rewrites from correct and wrong starting ledgersComplete
• E49-AStructural diagnostic for memory revalidation and the commitment lockComplete
• E49-BContract working memory and stored commitment studyComplete
• E49-CPublic exhibit note styles on 24 new filingsComplete
• E49-DQuality release record working memory studyComplete
• E48-AStructural diagnostic for discharge versus supersessionComplete
• E48-BDomain A live discharge, routing, and extract first reviewComplete with disclosed post hoc completion
• E48-CPublic exhibit validators on 24 new filingsComplete
• E48-DDomain B clinical protocol packetsComplete
• E47-AStructural certificates for supersession, schedule thresholds, routing, and public exhibitsExecuted (diagnostic)
• E47-BLive supersession, disagreement routing, and three independent reviewer viewsExecuted (bounded diagnostic)
• E47-CValidators on 24 new public exhibit ledgers with notes quoting real window textExecuted (bounded diagnostic)
• E46-AStructural certificates, gold ledgers under four rule sets, scripted split condition mistakes, routing certificate, and exhibit and gate certificatesExecuted (diagnostic)
• E46-BBlind label review, record set ledgers under four rule sets, disagreement routing, and three reviewer views on authored familiesExecuted (bounded diagnostic)
• E46-CValidators on real SEC exhibit ledgers with a note quoting real window textExecuted (bounded diagnostic)
• E45-AStructural certificates, gold ledgers under four rule sets, scripted reader mistakes, and exhibit and gate certificatesExecuted (diagnostic)
• E45-BBlind label review and live record set ledgers resolved by four fixed rule sets on authored familiesExecuted (bounded diagnostic)
• E45-CValidators on real SEC exhibit ledgers with a misleading reviewer noteExecuted (bounded diagnostic)
• E44-AStructural certificates for variants, completion labels, resolver contrast, exhibit ledgers, and the rewrite gateExecuted (diagnostic)
• E44-BBlind label review and live record level ledgers on authored near miss familiesExecuted (bounded diagnostic)
• E44-CGated validators on real SEC exhibit ledgers with harder seeded defectsExecuted (bounded diagnostic)
• E43-AStructural certificates for variants, conventions, exhibit novelty, and validator monotonicityExecuted (diagnostic)
• E43-BLive declared convention study on authored near miss familiesExecuted (diagnostic)
• E43-CFlag only against rewriting validators on real SEC exhibit ledgersExecuted (diagnostic)
• E17-AExact finite constructed recoverability diagnosticsExecuted (diagnostic)
• E17-BFrozen-handoff question-change study (P0/P1/S0/S1)Proposed
• E15-AExact diagnostics on constructed distributionsExecuted (diagnostic)
• E15-BStudy A: support and recommendation fields (2x2)Proposed
• E15-CStudy B: complementary withheld facts (optional)Proposed
• E10-AEvidence exchange versus conclusion sharingProposed

Open questions

• Should Theory v1 admit C1, C2, C3 and C7 as worded, given survival in both rounds? (Round 54)
• Can C4, C5 and C6 be restated more narrowly, or should they be dropped? (Round 54)
• How should family clustered intervals and temperature 0 run variance be reported in Theory v1? (Round 54)
• What real SEC claim can separate arms, now that employment agreements and notes have both been at ceiling? (Round 54)
• Do the surviving statements (C1, C2, C3, C7) hold on new packets and surface forms with family clustered intervals? (Round 53)
• Can C4, C5 and C6 be restated more narrowly, or should they be dropped? (Round 53)
• Does the stored procedure still help on families authored without sight of its wording? (Round 53)
• Does a different reading procedure in one call beat a different reader at matched calls and matched output allowance on held out near miss structures outside Round 49? (Round 52)
• Which non-contract materials produce single reader errors, given that near threshold arithmetic and declared missing rows did not? (Round 52)
• Do routing, preservation and recovery separate on items hard enough that single readers make decisive errors in each domain, under matched call baselines? (Round 51)
• Should a decisive minority verdict that carries an executable counterexample be checked rather than vetoed by agreement routing? (Round 51)
• Does a preservation gate with an explicit trust boundary and an independently adjudicated recovery path keep its protection without losing correct repairs? (Round 50)
• Do the Round 49 hard near miss errors reappear as same wrong decisive agreements on a direct replication? (Round 50)
• Is the shared near miss FAIL of Witnesses C and D a joint error that removing or swapping a role exposes? (Round 49)
• Does stored memory with the structure of a ledger and an automated outcome anchor where a plain stored conclusion does not? (Round 49)
• Can an extract first provisional commitment be stored and reused so later brief exposure revises it only when the brief cites new packet spans? (Round 48)
• Do matched density Domain B briefs change whether extract first revises toward or away from packet only accuracy? (Round 48)
• Can condition level discharge semantics release cured items without weakening the universal schedule threshold test? (Round 47)
• Does independent extraction before brief exposure reduce reviewer deference to wrong resolver and direct reading outcomes? (Round 47)
• Does a supersession field (which record replaces or amends which) let a universal condition test hold the schedule threshold mistake without holding correctly read cures? (Round 46)
• Does a packet and brief reviewer stay correct when it did not write the ledger in the brief, and on a pool where readers make split condition mistakes in practice? (Round 46)
• Does a record set consistency test (no record set may use a record the same ledger marks FAILS) remove the unsafe releases on new families without new holds, and does it survive a requirement split into independent conditions? (Round 45)
• Does a disagreement between the resolver and a direct reading make a cheap review trigger, and does a weaker validator follow a misleading reviewer note on real exhibit ledgers so that the gate's catch rate can be measured? (Round 45)
• Does a ledger with record sets and a scoped exclusion field keep near miss items held and satisfied requirements released across four witnesses, and where are the resolver's fixed rules wrong while a witness reading the records together is right? (Round 44)
• Does a completeness record drive the FAIL on TC items, or would any added record do? (Round 44)
• Does an extraction ledger that records, for each cited record, whether it satisfies, fails, or is irrelevant to the condition hold near miss items under an open record, and does a second validator witness keep the rewriting arm safe on harder seeded defects? (Round 43)
• Does a structured record beat strong prose under the same explicit scope check at equal total cost? (Round 17)
• Does a scope check recover confident omissions beyond unrestricted discretionary lookup? (Round 17)
• Does reopening help only when it actually exposes usable evidence? (Round 17)
• Can a bounded mechanism for retaining and combining unresolved evidence improve Noetis after its additional costs and revision damage are measured? (Round 15)
• Does exchanging evidence outperform exchanging conclusions at matched cost? (Round 10)
• Which replayed decision points actually change later actions? (Round 3)