KEEP: A KV-Cache-Centric Memory Management System for Efficient Embodied Planning

Memory-augmented Large Language Models (LLMs) have demonstrated remarkable capability for complex and long-horizon embodied planning. By keeping track of past experiences and environmental states, memory enables LLMs to maintain a global view, thereby avoiding repetitive exploration. However, existing approaches often store the memory as raw text, leading to excessively long prompts and high prefill latency. While it is possible to store and reuse the KV caches, the efficiency benefits are greatly undermined due to frequent KV cache updates. In this paper, we propose KEEP, a KV-cache-centric memory management system for efficient embodied planning. KEEP features 3 key innovations: (1) a Static-Dynamic Memory Construction algorithm that reduces KV cache recomputation by mixed-granularity memory group; (2) a Multi-hop Memory Re-computation algorithm that dynamically identifies important cross-attention among different memory groups and reconstructs memory interactions iteratively; (3) a Layer-balanced Memory Loading that eliminates unbalanced KV cache loading and cross-attention computation across different layers. Extensive experimental results have demonstrated that KEEP achieves 2.68x speedup with negligible accuracy loss compared with text-based memory methods on ALFRED dataset. Compared with the KV re-computation method CacheBlend (EuroSys'25), KEEP shows 4.13% success rate improvement and 1.90x time-to-first-token (TTFT) reduction. Our code is available on https://github.com/PKU-SEC-Lab/KEEP_Embodied_Memory.

DySL-VLA: Efficient Vision-Language-Action Model Inference via Dynamic-Static Layer-Skipping for Robot Manipulation

Vision-Language-Action (VLA) models have shown remarkable success in robotic tasks like manipulation by fusing a language model's reasoning with a vision model's 3D understanding. However, their high computational cost remains a major obstacle for real-world applications that require real-time performance. We observe that the actions within a task have varying levels of importance: critical steps demand high precision, while less important ones can tolerate more variance. Leveraging this insight, we propose DySL-VLA, a novel framework that addresses computational cost by dynamically skipping VLA layers based on each action's importance. DySL-VLA categorizes its layers into two types: informative layers, which are consistently executed, and incremental layers, which can be selectively skipped. To intelligently skip layers without sacrificing accuracy, we invent a prior-post skipping guidance mechanism to determine when to initiate layer-skipping. We also propose a skip-aware two-stage knowledge distillation algorithm to efficiently train a standard VLA into a DySL-VLA. Our experiments indicate that DySL-VLA achieves 2.1% improvement in success length over Deer-VLA on the Calvin dataset, while simultaneously reducing trainable parameters by a factor of 85.7 and providing a 3.75x speedup relative to the RoboFlamingo baseline at iso-accuracy. Our code is available on https://github.com/PKU-SEC-Lab/DYSL_VLA.

RoboGene: Boosting VLA Pre-training via Diversity-Driven Agentic Framework for Real-World Task Generation

The pursuit of general-purpose robotic manipulation is hindered by the scarcity of diverse, real-world interaction data. Unlike data collection from web in vision or language, robotic data collection is an active process incurring prohibitive physical costs. Consequently, automated task curation to maximize data value remains a critical yet under-explored challenge. Existing manual methods are unscalable and biased toward common tasks, while off-the-shelf foundation models often hallucinate physically infeasible instructions. To address this, we introduce RoboGene, an agentic framework designed to automate the generation of diverse, physically plausible manipulation tasks across single-arm, dual-arm, and mobile robots. RoboGene integrates three core components: diversity-driven sampling for broad task coverage, self-reflection mechanisms to enforce physical constraints, and human-in-the-loop refinement for continuous improvement. We conduct extensive quantitative analysis and large-scale real-world experiments, collecting datasets of 18k trajectories and introducing novel metrics to assess task quality, feasibility, and diversity. Results demonstrate that RoboGene significantly outperforms state-of-the-art foundation models (e.g., GPT-4o, Gemini 2.5 Pro). Furthermore, real-world experiments show that VLA models pre-trained with RoboGene achieve higher success rates and superior generalization, underscoring the importance of high-quality task generation. Our project is available at https://robogene-boost-vla.github.io.

RoboAug: One Annotation to Hundreds of Scenes via Region-Contrastive Data Augmentation for Robotic Manipulation

Enhancing the generalization capability of robotic learning to enable robots to operate effectively in diverse, unseen scenes is a fundamental and challenging problem. Existing approaches often depend on pretraining with large-scale data collection, which is labor-intensive and time-consuming, or on semantic data augmentation techniques that necessitate an impractical assumption of flawless upstream object detection in real-world scenarios. In this work, we propose RoboAug, a novel generative data augmentation framework that significantly minimizes the reliance on large-scale pretraining and the perfect visual recognition assumption by requiring only the bounding box annotation of a single image during training. Leveraging this minimal information, RoboAug employs pre-trained generative models for precise semantic data augmentation and integrates a plug-and-play region-contrastive loss to help models focus on task-relevant regions, thereby improving generalization and boosting task success rates. We conduct extensive real-world experiments on three robots, namely UR-5e, AgileX, and Tien Kung 2.0, spanning over 35k rollouts. Empirical results demonstrate that RoboAug significantly outperforms state-of-the-art data augmentation baselines. Specifically, when evaluating generalization capabilities in unseen scenes featuring diverse combinations of backgrounds, distractors, and lighting conditions, our method achieves substantial gains over the baseline without augmentation. The success rates increase from 0.09 to 0.47 on UR-5e, from 0.16 to 0.60 on AgileX, and from 0.19 to 0.67 on Tien Kung 2.0. These results highlight the superior generalization and effectiveness of RoboAug in real-world manipulation tasks. Our project is available at https://x-roboaug.github.io/.

Brain4FMs: A Benchmark of Foundation Models for Electrical Brain Signal

Brain Foundation Models (BFMs) are transforming neuroscience by enabling scalable and transferable learning from neural signals, advancing both clinical diagnostics and cutting-edge neuroscience exploration. Their emergence is powered by large-scale clinical recordings, particularly electroencephalography (EEG) and intracranial EEG, which provide rich temporal and spatial representations of brain dynamics. However, despite their rapid proliferation, the field lacks a unified understanding of existing methodologies and a standardized evaluation framework. To fill this gap, we map the benchmark design space along two axes: (i) from the model perspective, we organize BFMs under a self-supervised learning (SSL) taxonomy; and (ii) from the dataset perspective, we summarize common downstream tasks and curate representative public datasets across clinical and human-centric neurotechnology applications. Building on this consolidation, we introduce Brain4FMs, an open evaluation platform with plug-and-play interfaces that integrates 15 representative BFMs and 18 public datasets. It enables standardized comparisons and analysis of how pretraining data, SSL strategies, and architectures affect generalization and downstream performance, guiding more accurate and transferable BFMs. The code is available at https://anonymous.4open.science/r/Brain4FMs-85B8.

TEAM: Temporal-Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration

Diffusion large language models (dLLMs) have recently gained significant attention due to their inherent support for parallel decoding. Building on this paradigm, Mixture-of-Experts (MoE) dLLMs with autoregressive (AR) initialization have further demonstrated strong performance competitive with mainstream AR models. However, we identify a fundamental mismatch between MoE architectures and diffusion-based decoding. Specifically, a large number of experts are activated at each denoising step, while only a small subset of tokens is ultimately accepted, resulting in substantial inference overhead and limiting their deployment in latency-sensitive applications. In this work, we propose TEAM, a plug-and-play framework that accelerates MoE dLLMs by enabling more accepted tokens with fewer activated experts. TEAM is motivated by the observation that expert routing decisions exhibit strong temporal consistency across denoising levels as well as spatial consistency across token positions. Leveraging these properties, TEAM employs three complementary expert activation and decoding strategies, conservatively selecting necessary experts for decoded and masked tokens and simultaneously performing aggressive speculative exploration across multiple candidates. Experimental results demonstrate that TEAM achieves up to 2.2x speedup over vanilla MoE dLLM, with negligible performance degradation. Code is released at https://github.com/PKU-SEC-Lab/TEAM-MoE-dLLM.

HyPER: Bridging Exploration and Exploitation for Scalable LLM Reasoning with Hypothesis Path Expansion and Reduction

Scaling test-time compute with multi-path chain-of-thought improves reasoning accuracy, but its effectiveness depends critically on the exploration-exploitation trade-off. Existing approaches address this trade-off in rigid ways: tree-structured search hard-codes exploration through brittle expansion rules that interfere with post-trained reasoning, while parallel reasoning over-explores redundant hypothesis paths and relies on weak answer selection. Motivated by the observation that the optimal balance is phase-dependent and that correct and incorrect reasoning paths often diverge only at late stages, we reformulate test-time scaling as a dynamic expand-reduce control problem over a pool of hypotheses. We propose HyPER, a training-free online control policy for multi-path decoding in mixture-of-experts models that reallocates computation under a fixed budget using lightweight path statistics. HyPER consists of an online controller that transitions from exploration to exploitation as the hypothesis pool evolves, a token-level refinement mechanism that enables efficient generation-time exploitation without full-path resampling, and a length- and confidence-aware aggregation strategy for reliable answer-time exploitation. Experiments on four mixture-of-experts language models across diverse reasoning benchmarks show that HyPER consistently achieves a superior accuracy-compute trade-off, improving accuracy by 8 to 10 percent while reducing token usage by 25 to 40 percent.

Full end-to-end diagnostic workflow automation of 3D OCT via foundation model-driven AI for retinal diseases

Optical coherence tomography (OCT) has revolutionized retinal disease diagnosis with its high-resolution and three-dimensional imaging nature, yet its full diagnostic automation in clinical practices remains constrained by multi-stage workflows and conventional single-slice single-task AI models. We present Full-process OCT-based Clinical Utility System (FOCUS), a foundation model-driven framework enabling end-to-end automation of 3D OCT retinal disease diagnosis. FOCUS sequentially performs image quality assessment with EfficientNetV2-S, followed by abnormality detection and multi-disease classification using a fine-tuned Vision Foundation Model. Crucially, FOCUS leverages a unified adaptive aggregation method to intelligently integrate 2D slices-level predictions into comprehensive 3D patient-level diagnosis. Trained and tested on 3,300 patients (40,672 slices), and externally validated on 1,345 patients (18,498 slices) across four different-tier centers and diverse OCT devices, FOCUS achieved high F1 scores for quality assessment (99.01%), abnormally detection (97.46%), and patient-level diagnosis (94.39%). Real-world validation across centers also showed stable performance (F1: 90.22%-95.24%). In human-machine comparisons, FOCUS matched expert performance in abnormality detection (F1: 95.47% vs 90.91%) and multi-disease diagnosis (F1: 93.49% vs 91.35%), while demonstrating better efficiency. FOCUS automates the image-to-diagnosis pipeline, representing a critical advance towards unmanned ophthalmology with a validated blueprint for autonomous screening to enhance population scale retinal care accessibility and efficiency.

IntraSlice: Towards High-Performance Structural Pruning with Block-Intra PCA for LLMs

Large Language Models (LLMs) achieve strong performance across diverse tasks but face deployment challenges due to their massive size. Structured pruning offers acceleration benefits but leads to significant performance degradation. Recent PCA-based pruning methods have alleviated this issue by retaining key activation components, but are only applied between modules in order to fuse the transformation matrix, which introduces extra parameters and severely disrupts activation distributions due to residual connections. To address these issues, we propose IntraSlice, a framework that applies block-wise module-intra PCA compression pruning. By leveraging the structural characteristics of Transformer modules, we design an approximate PCA method whose transformation matrices can be fully fused into the model without additional parameters. We also introduce a PCA-based global pruning ratio estimator that further considers the distribution of compressed activations, building on conventional module importance. We validate our method on Llama2, Llama3, and Phi series across various language benchmarks. Experimental results demonstrate that our approach achieves superior compression performance compared to recent baselines at the same compression ratio or inference speed.

Attention Sink Forges Native MoE in Attention Layers: Sink-Aware Training to Address Head Collapse

Large Language Models (LLMs) often assign disproportionate attention to the first token, a phenomenon known as the attention sink. Several recent approaches aim to address this issue, including Sink Attention in GPT-OSS and Gated Attention in Qwen3-Next. However, a comprehensive analysis of the relationship among these attention mechanisms is lacking. In this work, we provide both theoretical and empirical evidence demonstrating that the sink in Vanilla Attention and Sink Attention naturally construct a Mixture-of-Experts (MoE) mechanism within attention layers. This insight explains the head collapse phenomenon observed in prior work, where only a fixed subset of attention heads contributes to generation. To mitigate head collapse, we propose a sink-aware training algorithm with an auxiliary load balancing loss designed for attention layers. Extensive experiments show that our method achieves effective head load balancing and improves model performance across Vanilla Attention, Sink Attention, and Gated Attention. We hope this study offers a new perspective on attention mechanisms and encourages further exploration of the inherent MoE structure within attention layers.