LLM-as-Jev: LLM의 Jev형 의사결정 능력 분석 및 파인튜닝

같은 소식, 다른 매체 · 8곳

• arXivFast Models, Slow Evidence: A Paired and Self-Audited Evaluation of System-1 Decision Models for LLM Agent Harnesses
• arXivAutoCompact: Learning When to Compact Context in Long-Horizon Coding Agents
• arXivAURAL: Adaptive Latent Reasoning with Joint Chunk for Speech Language Models
• arXivUniversal Byte-Level Encoding: UTF-8/UTF-16 Routing to Reduce Cross-Script Token-Budget Disparities
• arXivWhat Makes Something Hard(er)? Explaining Question Difficulty in Natural Language
• arXivGroundability, Not Scale Alone: When Weak Reviewers Can Audit Strong Coding Agents
• arXivCUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking
• arXivScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research

원문 발췌 · English

Jev-style decision models have gained surging traction by making decision-making dramatically faster and more efficient, returning categorical probability distributions over predefined options without generating free-form text. However, because Jev's architecture and training methodologies remain proprietary, how to replicate these capabilities effectively remains an open question. In this work, we investigate the extent to which general-purpose LLMs already possess this capability out of the box, and when fine-tuning is actually necessary. We present LLM-as-Jev, an architecture-preserving framework offering both a training-free inference recipe and a fine-tuning objective with KL-divergence anchors. Across Qwen3.5-4B and Qwen3-0.6B, we find that modern LLMs are inherently effective decision models: off the shelf, frozen models match or exceed existing training-free readouts on the same 4B backbone, scale to arbitrary option counts, and natively handle multimodal decisions. They also seamlessly inherit the backbone's thinking mode: while Jev was introduced as a System 1 model, "Thinking Jev" leverages budgeted test-time reasoning to boost accuracy on challenging tasks without training. Fine-tuning provides targeted gains for smaller models and complex routing, while our KL anchors prevent behavioral degradation in general text generation.