Standards & Methodology
How to read and evaluate the Chronicle
AI4Math Chronicle is not a news feed or paper database. It organizes public sources into stable historical event records and keeps historical significance, evidence strength, verification status, and formal assurance separate.
Four dimensions of the Chronicle trust model
How important is this historically?
H1–H3 describes the event’s durable place in AI4Math history or an important technical direction.
How mature is the public evidence?
Source tiers and E0–E4 track the progression from claims and research records to inspectable artifacts and independent checks.
How far is the core event description corroborated?
Verification records the Chronicle’s current event-level evidence state for what happened; it is not a verdict on ultimate mathematical truth.
How far has a formal artifact been checked?
Formal Assurance distinguishes no artifact, artifact availability, machine checking, and independent replay.
These four dimensions remain independent rather than collapsing into one “trust score.” Source tiers explain where evidence comes from; they do not replace significance, verification state, or mathematical correctness judgments.
1. What enters the Chronicle?
The Chronicle uses broad discovery and strict publication gates: candidates may come from papers, formal proof artifacts, lab announcements, GitHub, scholarly discussion, or community signals, but only historically or field-significant canonical events enter the published record.
In scope are historically meaningful discoveries, proofs, formalizations, competition results, systems, benchmarks, datasets, infrastructure, field-building milestones, and controversies. Field building covers durable milestones in the formation, institutionalization, scholarly norms, or governance of AI4Math—for example a sustained dedicated journal or conference, or a community statement that materially changes research norms. Ordinary petitions, one-off workshops, routine special issues, and ephemeral community activity are excluded by default, as are routine model launches, ordinary paper flow, incremental benchmark SOTA, pure prompt engineering, routine formalization updates, and marketing-only releases.
For a purely field-building event, technical dimensions that genuinely do not apply use not_applicable rather than misclassifying the record as tooling_only, unclear, or informal. not_applicable is restricted to events that include field_building.
2. How should historical significance be read?
| H1 | Historical Milestone: a durable turning point in AI for Mathematics, such as a major new result, proof, disproof, landmark formalization, or repeated research-level breakthrough. Final designation requires human review. |
|---|---|
| H2 | Field Milestone: clear significance to an important AI4Math subfield, technical direction, or field-building process, including math-specific systems, benchmarks, datasets, proof search, formalization infrastructure, important competition results, durable venues, or norm-setting milestones. |
| H3 | Context Event: primarily establishes historical context rather than representing a substantive AI4Math milestone. |
3. How does the Chronicle organize evidence?
The Chronicle does not compress trust into one score. It records source quality, evidence maturity, current verification status, and formal-artifact assurance as separate dimensions.
Source hierarchy (S1–S5)
| S1 | Primary research sources such as papers, journals/conferences, formal proof artifacts, official research repositories, and dataset artifacts. |
|---|---|
| S2 | Official announcements and formal records from labs, universities, researchers, scholarly venues, or related institutions. |
| S3 | Independent scholarly corroboration or analysis, including reproduction, external scoring, peer review, or artifact replay. |
| S4 | Reputable science and technology media. |
| S5 | Community discovery signals such as HN, Reddit, X, or forums; useful for discovery but insufficient alone for major research claims or field-building milestones. |
Evidence level (E0–E4)
| E0 | Unsubstantiated: insufficient public evidence is recorded. |
|---|---|
| E1 | Primary Claim: the originating party has made a public claim. |
| E2 | Research Record: a public research record exists. |
| E3 | Inspectable Artifact: code, proof, data, or another inspectable artifact is available. |
| E4 | Independent Check: independent checking or corroboration is recorded. |
Verification status and formal assurance
Verification Status describes the current evidence state for what happened, such as claimed, paper_released, under_verification, partially_verified, independently_verified, disputed, corrected, or retracted. Formal Assurance separately describes formal-artifact assurance: none, artifact_available, machine_checked, or independently_replayed. For field-building events, verification concerns whether the statement, venue, governance, or institutional event occurred as recorded; it does not imply Chronicle endorsement of a normative position or scholarly prestige.
4. How are uncertainty and disputes handled?
Chronicle verification concerns the event and its public evidence: what the originating source claimed, which artifacts were released, and how far reliable independent sources corroborate the core event description. Verification is not a final adjudication of mathematical truth, historical priority, or attribution disputes.
`under_verification` marks material aspects still under external scrutiny; `partially_verified` records important corroboration with material questions still open; `independently_verified` means reliable independent sources substantively corroborate the core event description; `disputed`, `corrected`, and `retracted` preserve material contestation, correction, or withdrawal.
5. What happens when a record is wrong or incomplete?
Chronicle uses stable Event IDs and preserves auditable revision trails through verification history and corrections. When facts change, the canonical YAML should be updated rather than silently rewriting history. Corrections, source leads, missing-event proposals, and website problems enter review through public, traceable GitHub workflows.
This page explains why the Chronicle is correctable and how revision fits the trust model. Project orientation, usage paths, and practical submission actions are centralized on About so the standards page does not duplicate them.