The authors define trust as willingness to accept vulnerability to another party’s actions, while avoiding claims that models experience trust subjectively.
Activation steering shifts how language models decide to trust users
Across six instruction-tuned models, learned activation interventions changed reliance on unverifiable user claims in both directions.
AI Startup
Théo Lasnier · Romain Froger · Maxence Lasbordes · Djamé Seddah
Inria Paris · Sorbonne Université · Meta SuperIntelligence Labs · LightOn
Research Digest··3 min read
Lasnier and colleagues study trust from the assistant’s perspective: whether a model acts on claims whose accuracy or intent it cannot verify.
Why this paper
From LightOn and 3 others
In one line
Steering model activations causally controls how LLM assistants trust users, affecting safety behaviors.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§