Skip to content

ProbeBench subdirectory — register the 5 v0.0.1 reference probes #5

Open
Open
ProbeBench subdirectory — register the 5 v0.0.1 reference probes#5
Labels
atlasPublic registry of mech-interp publicationsprobebenchProbe leaderboard / categorical evaluation
@caiovicentino

Description

@caiovicentino
caiovicentino
opened on May 11, 2026

probebench/<sha10>.json for entries that map to ProbeBench leaderboard categories. Drives openinterp.org/probebench cross-linking.

FG/RG/CoTGuard/DeceptionGuard/EvalAwarenessGuard from probebench-data.ts should each get a registry entry with probebench_category field.

Metadata

Metadata

Assignees

No one assigned

    Labels

    atlasPublic registry of mech-interp publicationsprobebenchProbe leaderboard / categorical evaluation

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    OpenInterp Q2 2026
    Status
    Todo

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions