Intertemporal Preference Steering in Qwen3 via Contrastive Activation Addition
Quick summary
arXiv:2608.03892v1 Announce Type: new Abstract: We study linear representations of temporal horizon in the large language model Qwen3-32B and use them to change the model's time-related preferences, recommendations, and capabilities. We train contrastive linear probes on teacher-forced temporal-choice answers to find a short-term versus long-term direction in the model's residual stream, and evaluate contrastive activation-addition steering on a held-out binary temporal-choice task, an out-of-distribution monetary intertemporal-choice task, and a TravelPlanner capability benchmark. The central
Key takeaways
- arXiv:2608.03892v1 Announce Type: new Abstract: We study linear representations of temporal horizon in the large language model Qwen3-32B and use them to change the model's time-related preferences, recommendations, and capabilities.
- We train contrastive linear probes on teacher-forced temporal-choice answers to find a short-term versus long-term direction in the model's residual stream, and evaluate contrastive activation-addition steering on a held-out binary temporal-choice task, an out-of-distribution monetary intertemporal-choice task, and a TravelPlanner capability benchmark.
Why it matters
“Intertemporal Preference Steering in Qwen3 via Contrastive Activation Addition” highlights the need for repeatable measurement rather than a single impressive demonstration. Independent validation across datasets and clearly stated limitations determine whether a result can guide product decisions.
Member comments