Zhimu Zhou

Jian Zhao, Runze Liu, Kaiyan Zhang et al.

Recent advancements in Large Language Models (LLMs) have shown that it is promising to utilize Process Reward Models (PRMs) as verifiers to enhance the performance of LLMs. However, current PRMs face three key challenges: (1) limited process supervision and generalization capabilities, (2) dependence on scalar value prediction without leveraging the generative abilities of LLMs, and (3) inability to scale the test-time compute of PRMs. In this work, we introduce GenPRM, a generative process reward model that performs explicit Chain-of-Thought (CoT) reasoning with code verification before providing judgment for each reasoning step. To obtain high-quality process supervision labels and rationale data, we propose Relative Progress Estimation (RPE) and a rationale synthesis framework that incorporates code verification. Experimental results on ProcessBench and several mathematical reasoning tasks show that GenPRM significantly outperforms prior PRMs with only 23K training data from MATH dataset. Through test-time scaling, a 1.5B GenPRM outperforms GPT-4o, and a 7B GenPRM surpasses Qwen2.5-Math-PRM-72B on ProcessBench. Additionally, GenPRM demonstrates strong abilities to serve as a critic model for policy model refinement. This work establishes a new paradigm for process supervision that bridges the gap between PRMs and critic models in LLMs. Our code, model, and data will be available in https://ryanliu112.github.io/GenPRM.

Zhimu Zhou, Yanpeng Zhao, Qiuyu Liao et al.

Visual planning represents a crucial facet of human intelligence, especially in tasks that require complex spatial reasoning and navigation. Yet, in machine learning, this inherently visual problem is often tackled through a verbal-centric lens. While recent research demonstrates the promise of fully visual approaches, they suffer from significant computational inefficiency due to the step-by-step planning-by-generation paradigm. In this work, we present EAR, an editing-as-reasoning paradigm that reformulates visual planning as a single-step image transformation. To isolate intrinsic reasoning from visual recognition, we employ abstract puzzles as probing tasks and introduce AMAZE, a procedurally generated dataset that features the classical Maze and Queen problems, covering distinct, complementary forms of visual planning. The abstract nature of AMAZE also facilitates automatic evaluation of autoregressive and diffusion-based models in terms of both pixel-wise fidelity and logical validity. We assess leading proprietary and open-source editing models. The results show that they all struggle in the zero-shot setting, finetuning on basic scales enables remarkable generalization to larger in-domain scales and out-of-domain scales and geometries. However, our best model that runs on high-end hardware fails to match the zero-shot efficiency of human solvers, highlighting a persistent gap in neural visual reasoning.

Chenghao Liu, Jiachen Zhang, Chengxuan Li et al.

Vision-Language-Action (VLA) models process visual inputs independently at each timestep, discarding valuable temporal information inherent in robotic manipulation tasks. This frame-by-frame processing makes models vulnerable to visual noise while ignoring the substantial coherence between consecutive frames in manipulation sequences. We propose Temporal Token Fusion (TTF), a training-free approach that intelligently integrates historical and current visual representations to enhance VLA inference quality. Our method employs dual-dimension detection combining efficient grayscale pixel difference analysis with attention-based semantic relevance assessment, enabling selective temporal token fusion through hard fusion strategies and keyframe anchoring to prevent error accumulation. Comprehensive experiments across LIBERO, SimplerEnv, and real robot tasks demonstrate consistent improvements: 4.0 percentage points average on LIBERO (72.4\% vs 68.4\% baseline), cross-environment validation on SimplerEnv (4.8\% relative improvement), and 8.7\% relative improvement on real robot tasks. Our approach proves model-agnostic, working across OpenVLA and VLA-Cache architectures. Notably, TTF reveals that selective Query matrix reuse in attention mechanisms enhances rather than compromises performance, suggesting promising directions for direct KQV matrix reuse strategies that achieve computational acceleration while improving task success rates.

Chenghao Liu, Zhimu Zhou, Jiachen Zhang et al.

Vision-and-Language Navigation (VLN) requires an agent to interpret natural language instructions and navigate complex environments. Current approaches often adopt a "black-box" paradigm, where a single Large Language Model (LLM) makes end-to-end decisions. However, it is plagued by critical vulnerabilities, including poor spatial reasoning, weak cross-modal grounding, and memory overload in long-horizon tasks. To systematically address these issues, we propose Memory Spatial Navigation(MSNav), a framework that fuses three modules into a synergistic architecture, which transforms fragile inference into a robust, integrated intelligence. MSNav integrates three modules: Memory Module, a dynamic map memory module that tackles memory overload through selective node pruning, enhancing long-range exploration; Spatial Module, a module for spatial reasoning and object relationship inference that improves endpoint recognition; and Decision Module, a module using LLM-based path planning to execute robust actions. Powering Spatial Module, we also introduce an Instruction-Object-Space (I-O-S) dataset and fine-tune the Qwen3-4B model into Qwen-Spatial (Qwen-Sp), which outperforms leading commercial LLMs in object list extraction, achieving higher F1 and NDCG scores on the I-O-S test set. Extensive experiments on the Room-to-Room (R2R) and REVERIE datasets demonstrate MSNav's state-of-the-art performance with significant improvements in Success Rate (SR) and Success weighted by Path Length (SPL).