Jev, the yes/no-only model, sparks debate: agents need judgment parts, not bigger engines
冷技术热思考 · wechat · 2026-09-23
Jev, a model that can only answer multiple-choice, scoring and yes/no questions with calibrated probabilities — no chatting, coding or vision — has stirred debate in the AI community. Its fixed answer space makes it fast and cheap.
Author Wang Yuan (ex-NetEase VP, founder of remio) frames it with a car metaphor: the LLM is the engine, context engineering is refining, and Jev is a sensor/ECU-class "judgment part." Emerging use cases include BrowserUse (picking which numbered element to click, cutting a Google Flights search from 9+ to 7 seconds), coding agents (filtering irrelevant tool output, routing easy tasks to cheap models, pre-screening risky code changes), and massive per-item classification (paper topics, spam, fraud, training-data curation).
Community tests show 7B models can be more accurate; Jev's value is turning judgment from "writing essays" into "filling answer sheets." The author plans to wire Jev into remio's retrieval filtering, memory arbitration and automation, under one principle: wherever a judgment suffices, don't generate.
Related event: Decision-Only Model Jev Goes Viral as a Cheaper DeepSeek-Style Threat(3 posts)→
More from coding & agent
- Loom founder launches Roma 2.0, a shared to-do list that orchestrates AI agents — thetripathi58 · 2026-11-12
- Sosumi Eval: Apple Docs Access Lifts Coding Agent Pass Rates from ~33% to 94-98% — JordanMorgan10 · 2026-11-12
- CypherGuard fixes structurally wrong LLM Cypher queries, hitting 88.5% on CypherBench — JeremyCMorgan · 2026-11-12
- Building an End-to-End Hedge Fund in Python with Claude Code: A 5-Step Workflow — mdancho84 · 2026-11-12
- Jeff Dean: comfortable with engineers spending up to $10M/year on AI tools — deedydas · 2026-11-12
- Developer builds VladOS, a personal life dashboard, by just describing it to an AI agent — haltakov · 2026-11-12