A Causal Model for Locating and Unlocking Sandbagging in Model Organisms

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

arXiv:2608.29461v1 Announce Type: new Abstract: Sandbagging models strategically underperform on evaluations while retaining the capabilities being measured. The evaluations that guide frontier-model deployment and governance then understate what these models can do. To understand the mechanism, we propose a causal model of how sandbagging is carried in the residual stream. Early layers write the sandbagging intent onto a single axis of the stream, and a later layer reads that axis and commits the answer. We study three instruction-tuned models (Qwen2.5-7B, Llama-3-8B, and Mistral-7B) and four ways of installing a sandbagging lock (prompting, fine-tuning, reinforcement learning, and circuit breaking). Each lock decides from the prompt whether to sandbag, and the fine-tuned and circuit-broken locks answer honestly whenever a password appears. The causal model predicts a window of layers, after the last sandbagging write and before the answer commit, in which a single-layer reference graft of the sandbagging axis to its honest value restores the full capability. The single-layer graft recovers the capability in 28 of the 33 runs of the prompted, fine-tuned, and RL-trained locks, with a median held-out recovery of 96%. The circuit-broken lock rewrites the whole state through a band of layers, and the single-layer graft fails at every layer. We therefore introduce a second intervention, context grafting, which replays the password's cached key/value activations so that every layer's attention reads them as additional context. Context grafting provably and empirically restores …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://arxiv.org/abs/2608.29461
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.