paper-with-me

홈 › Papers

interwhen: A Generalizable Framework for Steering Reasoning Models with Test-time Verification

2026-02-05 · Vishak K Bhat, Prateek Chanda, Vijval Ekbote, Ashmit Khandelwal, Maitreyi Swaroop, Vineeth N. Balasubramanian, Subbarao Kambhampati, Nagarajan Natarajan, Amit Sharma arxiv

Reasoning models produce long traces of intermediate decisions and tool calls, making test-time verification important for ensuring correctness. Existing approaches either verify only the final answer, which misses early errors, or rely on branch-and-verify strategies that explore multiple trajectories. We introduce interwhen, a single-trajectory verification framework that steers model behavior by providing feedback on intermediate reasoning traces. It addresses two key challenges. First, given a set of verifiers, obtaining verifiable states from the reasoning trace typically requires prompt engineering or external task decomposition into fixed steps. Instead, we propose a monitoring system that periodically polls the reasoning trace and forks inference of the reasoning model to recover intermediate states. Verifiers are run asynchronously alongside generation, adding negligible overhead on correct executions and intervening only when violations occur. Second, beyond math and code, a central challenge for process verification is the scarcity of verifiers. interwhen addresses this through automatic verifier synthesis from natural-language policy documents. Given a policy, it can generate code-based verifiers, including provably correct verifiers in Lean and z3. Together, these contributions yield a plug-and-play test-time verification system that can improve task completion and policy compliance of any reasoning agent. On reasoning benchmarks where policies encode mathematical or logical constraints, interwhen achieves near-perfect accuracy for reasoning models using a fraction of the tokens of baselines. On agentic benchmarks with policy-based verifier generation, it enables improvements in task quality for SLMs without any finetuning, e.g., task completion rate of Qwen3-30B jumps from 32% to 87% on the telecom domain in tau2-bench. Code at https://github.com/microsoft/interwhen.

📄 PDF Abstract BibTeX arXiv:2602.11202

Code (0)

등록된 구현이 없습니다.

Tasks

Prompt Engineering

Similar Papers 제목 키워드 기반

ATLAS: Verifier-Guided Adaptive Latent Activation Steering for Efficient LLM Reasoning

2026-01-06 · Tuc Nguyen, Thai Le arxiv

Recent work on activation and latent steering has demonstrated that modifying internal representations can effectively guide large language models (LLMs) toward improved reasoning and efficiency without updating model pa…

Forecasting With LLMs: Improved Generalization Through Feature Steering

2026-06-25 · Humzah Merchant, Bradford Levy arxiv

Successful forecasting involves identifying patterns between historical and future states of the world which generalize to future observations. We apply LLMs to a variety of forecasting tasks and inspect their internal s…

Test-time Diverse Reasoning by Riemannian Activation Steering

2025-11-11 · Ly Tran Ho Khanh, Dongxuan Zhu, Man-Chung Yue, Viet Anh Nguyen arxiv

Best-of-$N$ reasoning improves the accuracy of language models in solving complex tasks by sampling multiple candidate solutions and then selecting the best one based on some criteria. A critical bottleneck for this stra…

MoralReason: Generalizable Moral Decision Alignment For LLM Agents Using Reasoning-Level Reinforcement Learning

2025-11-15 · Zhiyu An, Wan Du arxiv

Large language models are increasingly influencing human moral decisions, yet current approaches focus primarily on evaluating rather than actively steering their moral decisions. We formulate this as an out-of-distribut…

Reinforcement LearningMoral Scenarios

From Foresight to Forethought: VLM-In-the-Loop Policy Steering via Latent Alignment

2025-02-03 · Yilin Wu, Ran Tian, Gokul Swamy, Andrea Bajcsy

While generative robot policies have demonstrated significant potential in learning complex, multimodal behaviors from demonstrations, they still exhibit diverse failures at deployment-time. Policy steering offers an ele…