[2604.03962] Predict, Don't React: Value-Based Safety Forecasting for LLM Streaming

Computer Science > Computation and Language arXiv:2604.03962 (cs) [Submitted on 5 Apr 2026] Title:Predict, Don't React: Value-Based Safety Forecasting for LLM Streaming Authors:Pride Kavumba, Koki Wataoka, Huy H. Nguyen, Jiaxuan Li, Masaya Ohagi View a PDF of the paper titled Predict, Don't React: Value-Based Safety Forecasting for LLM Streaming, by Pride Kavumba and 4 other authors View PDF HTML (experimental) Abstract:In many practical LLM deployments, a single guardrail is used for both prompt and response moderation. Prompt moderation operates on fully observed text, whereas streaming response moderation requires safety decisions to be made over partial generations. Existing text-based streaming guardrails commonly frame this output-side problem as boundary detection, training models to identify the earliest prefix at which a response has already become unsafe. In this work, we introduce StreamGuard, a unified model-agnostic streaming guardrail that instead formulates moderation as a forecasting problem: given a partial prefix, the model predicts the expected harmfulness of likely future continuations. We supervise this prediction using Monte Carlo rollouts, which enables early intervention without requiring exact token-level boundary annotations. Across standard safety benchmarks, StreamGuard performs strongly both for input moderation and for streaming output moderation. At the 8B scale, StreamGuard improves aggregated input-moderation F1 from 86.7 to 88.2 and aggreg...