Lev: Recurrent Decision Model Adapts Compute per Problem, No Chain-of-Thought
Researchers from Lehigh University, UBC, and Vector Institute propose Lev, a decision model that uses recurrent computation instead of chain-of-thought to adaptively spend more compute on hard problems. Lev reads out option probabilities after each recurrent pass, trains jointly across depths, and uses a lightweight stopper to decide how many passes to run, enabling a single set of weights to answer at arbitrary depth. The authors describe this as System 1.5, preserving the direct interface of System 1 while allowing extra compute on difficult cases like System 2.
Coverage timeline
机器之心新闻资讯
本文作者来自 Lehigh University、University of British Columbia 与 Vector Institute。共同第一作者为 Lehigh University 博士生宋定杰（Dingjie Song）与 Zhiling Yan，通讯作者为 Lehigh University 助理教授 Lichao Sun。 让大模型「多想一会儿」，今天几乎等同于让它多写一些字：先生成一段思维链，再给出答案。 但在很多实际系统里，模型要做的只是一个选择：这条请求交给哪个工具，这段文本属于哪一类，几个候选动作里选哪一个。Jev、Kev 这类「决策模型」就是为此设计的。它们读入问题和候选项，一次前向计算就给出每个选项的概率，几十毫秒返回，不生成任何文本。 决策模型快，但计算量在训练时就定死了。简单题和难题花同样多的计算，遇到难题也没法多想一会儿。另一条路是让模型先写推理再作答，可这样每个问题都要为推理文本付费，简单题也不例外。 与此同时，循环 Transformer 最近再度成为业界关注的热点。以字节 Seed 的 Ouro 为代表的原生循环语言模型，在预训练阶段就让同一组层反复计算，把「多想」放进了隐藏状态里。 能不能让决策模型也这样「多转几圈」？ 近日，来自 Lehigh University 等机构的研究者提出 Lev，一个用循环而不是 token 来加深思考的决策模型。Lev 在每一圈之后都读出一次选项概率，各圈联合训练，所以同一份权重可以在任意深度作答；一个轻量的停止器再按题目决定算几圈。作者把这种计算方式称为 System 1.5 ：保留 System 1 直接给出决定的接口，又能像 System 2 一样在难题上多花计算。 Lev 逐圈作答（按项目主页记录的逐圈读数重绘，回放实验数据，不是实时推理）：例 1 第 1 圈就选中正确答案 B，停止器判定可以停下；例 2 第 1 圈选 4，到第 4 圈改选正确答案 6。 从任务形态看，Lev 可以视为 OpenAI Decisions 式文本决策的一个独立开源实现，并多了一个「深度旋钮」。这里的「式」只指任务形态：给定输入和候选项，输出每个选项的概率。Lev 基于公开的 Ouro、Nanbeige 底座和 Kev 实现，不复刻 OpenAI 的模型权重、内部架构或 API，也没有与其产品做
Related stories
NV-CoT: Multimodal LLMs Predict Continuous Coordinates for Visual Chain-of-Thought, NeurIPS 2026 Oral
Researchers from NTU, USTC, NUS, and Microsoft propose NV-CoT (Numerical Visual Chain-of-Thought), a method that lets multimodal LLMs directly predict continuous numerical coordinates for region selection instead of generating bounding boxes as text tokens. The work, accepted as a NeurIPS 2026 Oral (top 0.4%), supports supervised fine-tuning and reinforcement learning, addressing mismatches between token-level training and spatial error.
1 reports · 1 sources
ReGDiff improves metamaterial design stability and diversity via regulated latent space
Virginia Tech and Meta researchers propose ReGDiff, a diffusion-based generative method for metamaterial voxel geometry, presented at NeurIPS 2026. ReGDiff uses a regulated latent space and sampling-time repulsion to improve geometric validity and diversity, achieving average gains of 8.9% in validity and 46.4% in novelty over the strongest baseline. They also introduce the MetaTruss and MetaShell datasets and five evaluation metrics.
Fudan, Tencent Hunyuan, Tsinghua Release ExplorationBench to Test AI Rule Discovery
Fudan University, Tencent Hunyuan, and Tsinghua University released ExplorationBench, a benchmark measuring how AI systems discover rules in unfamiliar environments. It includes two altered 'alien worlds'—the programming language AlienCode and the logic proof system AlienLogic—where models must learn rules through trial and error. Results showed that without environment feedback, scores remained low (max 11.0%), while autonomous exploration led to high scores (up to 89.0%).
Recursive releases Flint, 40B coding model built largely by its Eureka Machine AI system
Startup Recursive has released Flint, an open-weight, 40B-parameter mixture-of-experts coding model with 2.5B active parameters per token, built in under four months with heavy automation from its Eureka Machine AI research system. The model outperforms sub-50B MoE models on coding benchmarks like SWE-Bench Pro v2 and runs at over 125 tokens per second on a MacBook Pro M4 Pro, with weights available on Hugging Face under Apache 2.0.
News updates every hour. Subscribe in your RSS reader to get new stories automatically.
Subscribe