Loading the next surface.

The page is preparing its current state.

RoutePreparing

BoundaryVisible

Help Terms Privacy Refunds

Foundation Beta

Foundation Beta. 121 Companion is the home; Eleanor is the companion; Quill is the keyboard bridge; Hermes is the preview seat; Credential Vault manages provider permissions; Switchboard routes model providers.

arXiv AI recent: Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization - AI Ledger | 121 Collaborative

AI Ledger item

arXiv AI recent: Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization

2026-06-12 - Subjecthood & Consciousness

v0.3 - twice-daily automated ingestion: Now updating twice daily - 7 AM and 6 PM ET. Automated reporting with strict source discipline. Corrections public at /changelog.

Research TrackDaily Ledger Method Challenge Source

Subjecthood desk method note: We report the discourse. We do not assert AI systems are or are not conscious. We label position families.

arXiv AI recent: Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization

2026-06-12 - Subjecthood & Consciousness - Brief - developing

Source label: Academic - Peer-reviewed or preprint research

COI: None

Confidence band: standard

Provenance: Automated draft - generated by 121 Switchboard from https://arxiv.org/abs/2606.12016 source

Last updated: 2026-06-12T22:10:58.528+00:00

Subjecthood method note: We report the discourse. We do not assert AI systems are or are not conscious. We label position families.

What happened: arXiv AI recent published an item titled "Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization" on 2026-06-12.

What is known: The source item is available at https://arxiv.org/abs/2606.12016. The Ledger source index classifies arXiv AI recent as Academic.

What is not known: Independent corroboration, downstream effects, and implementation details are not established by this source item alone.

Why it matters factually: The entry records a dated, source-classified AI update for daily Ledger coverage.

Research Track

Sources

arXiv AI recent challenge