Daily Papers

Stable Diffusion Models (SDMs) have shown remarkable proficiency in image synthesis. However, their broad application is impeded by their large model sizes and intensive computational requirements, which typically require expensive cloud servers for deployment. On the flip side, while there are many compact models tailored for edge devices that can reduce these demands, they often compromise on semantic integrity and visual quality when compared to full-sized SDMs. To bridge this gap, we introduce Hybrid SD, an innovative, training-free SDMs inference framework designed for edge-cloud collaborative inference. Hybrid SD distributes the early steps of the diffusion process to the large models deployed on cloud servers, enhancing semantic planning. Furthermore, small efficient models deployed on edge devices can be integrated for refining visual details in the later stages. Acknowledging the diversity of edge devices with differing computational and storage capacities, we employ structural pruning to the SDMs U-Net and train a lightweight VAE. Empirical evaluations demonstrate that our compressed models achieve state-of-the-art parameter efficiency (225.8M) on edge devices with competitive image quality. Additionally, Hybrid SD reduces the cloud cost by 66% with edge-cloud collaborative inference.

We introduce Parallel Coordinated Reasoning (PaCoRe), a training-and-inference framework designed to overcome a central limitation of contemporary language models: their inability to scale test-time compute (TTC) far beyond sequential reasoning under a fixed context window. PaCoRe departs from the traditional sequential paradigm by driving TTC through massive parallel exploration coordinated via a message-passing architecture in multiple rounds. Each round launches many parallel reasoning trajectories, compacts their findings into context-bounded messages, and synthesizes these messages to guide the next round and ultimately produce the final answer. Trained end-to-end with large-scale, outcome-based reinforcement learning, the model masters the synthesis abilities required by PaCoRe and scales to multi-million-token effective TTC without exceeding context limits. The approach yields strong improvements across diverse domains, and notably pushes reasoning beyond frontier systems in mathematics: an 8B model reaches 94.5% on HMMT 2025, surpassing GPT-5's 93.2% by scaling effective TTC to roughly two million tokens. We open-source model checkpoints, training data, and the full inference pipeline to accelerate follow-up work.

We introduce xLLM, an intelligent and efficient Large Language Model (LLM) inference framework designed for high-performance, large-scale enterprise-grade serving, with deep optimizations for diverse AI accelerators. To address these challenges, xLLM builds a novel decoupled service-engine architecture. At the service layer, xLLM-Service features an intelligent scheduling module that efficiently processes multimodal requests and co-locates online and offline tasks through unified elastic scheduling to maximize cluster utilization. This module also relies on a workload-adaptive dynamic Prefill-Decode (PD) disaggregation policy and a novel Encode-Prefill-Decode (EPD) disaggregation policy designed for multimodal inputs. Furthermore, it incorporates a distributed architecture to provide global KV Cache management and robust fault-tolerant capabilities for high availability. At the engine layer, xLLM-Engine co-optimizes system and algorithm designs to fully saturate computing resources. This is achieved through comprehensive multi-layer execution pipeline optimizations, an adaptive graph mode and an xTensor memory management. xLLM-Engine also further integrates algorithmic enhancements such as optimized speculative decoding and dynamic EPLB, collectively serving to substantially boost throughput and inference efficiency. Extensive evaluations demonstrate that xLLM delivers significantly superior performance and resource efficiency. Under identical TPOT constraints, xLLM achieves throughput up to 1.7x that of MindIE and 2.2x that of vLLM-Ascend with Qwen-series models, while maintaining an average throughput of 1.7x that of MindIE with Deepseek-series models. xLLM framework is publicly available at https://github.com/jd-opensource/xllm and https://github.com/jd-opensource/xllm-service.

This paper introduces VideoScan, an efficient vision-language model (VLM) inference framework designed for real-time video interaction that effectively comprehends and retains streamed video inputs while delivering rapid and accurate responses. A longstanding challenge in video understanding--particularly for long-term or real-time applications--stems from the substantial computational overhead caused by the extensive length of visual tokens. To address this, VideoScan employs a single semantic carrier token to represent each frame, progressively reducing computational and memory overhead during its two-phase inference process: prefilling and decoding. The embedding of the semantic carrier token is derived from an optimized aggregation of frame-level visual features, ensuring compact yet semantically rich representations. Critically, the corresponding key-value pairs are trained to retain contextual semantics from prior frames, enabling efficient memory management without sacrificing temporal coherence. During inference, the visual tokens of each frame are processed only once during the prefilling phase and subsequently discarded in the decoding stage, eliminating redundant computations. This design ensures efficient VLM inference even under stringent real-time constraints. Comprehensive experiments on diverse offline and online benchmarks demonstrate that LLaVA-Video, supported by our method, achieves up to sim 5times and 1.29times speedups compared to its original version and previous efficient streaming video understanding approaches, respectively. Crucially, these improvements are attained while maintaining competitive performance and ensuring stable GPU memory consumption (consistently sim 18GB, independent of video duration).

The efficiency of large vision-language models (LVLMs) is constrained by the computational bottleneck of the attention mechanism during the prefill phase and the memory bottleneck of fetching the key-value (KV) cache in the decoding phase, particularly in scenarios involving high-resolution images or videos. Visual content often exhibits substantial redundancy, resulting in highly sparse attention maps within LVLMs. This sparsity can be leveraged to accelerate attention computation or compress the KV cache through various approaches. However, most studies focus on addressing only one of these bottlenecks and do not adequately support dynamic adjustment of sparsity concerning distinct layers or tasks. In this paper, we present ZipVL, an efficient inference framework designed for LVLMs that resolves both computation and memory bottlenecks through a dynamic ratio allocation strategy of important tokens. This ratio is adaptively determined based on the layer-specific distribution of attention scores, rather than fixed hyper-parameters, thereby improving efficiency for less complex tasks while maintaining high performance for more challenging ones. Then we select important tokens based on their normalized attention scores and perform attention mechanism solely on those important tokens to accelerate the prefill phase. To mitigate the memory bottleneck in the decoding phase, we employ mixed-precision quantization to the KV cache, where high-bit quantization is used for caches of important tokens, while low-bit quantization is applied to those of less importance. Our experiments demonstrate that ZipVL can accelerate the prefill phase by 2.6times and reduce GPU memory usage by 50.0%, with a minimal accuracy reduction of only 0.2% on Video-MME benchmark over LongVA-7B model, effectively enhancing the generation efficiency of LVLMs.

In this technical report, we present SWE-Master, an open-source and fully reproducible post-training framework for building effective software engineering agents. SWE-Master systematically explores the complete agent development pipeline, including teacher-trajectory synthesis and data curation, long-horizon SFT, RL with real execution feedback, and inference framework design. Starting from an open-source base model with limited initial SWE capability, SWE-Master demonstrates how systematical optimization method can elicit strong long-horizon SWE task solving abilities. We evaluate SWE-Master on SWE-bench Verified, a standard benchmark for realistic software engineering tasks. Under identical experimental settings, our approach achieves a resolve rate of 61.4\% with Qwen2.5-Coder-32B, substantially outperforming existing open-source baselines. By further incorporating test-time scaling~(TTS) with LLM-based environment feedback, SWE-Master reaches 70.8\% at TTS@8, demonstrating a strong performance potential. SWE-Master provides a practical and transparent foundation for advancing reproducible research on software engineering agents. The code is available at https://github.com/RUCAIBox/SWE-Master.

Memory-augmented Large Language Models (LLMs) are essential for developing capable, long-term AI agents. Recently, applying Reinforcement Learning (RL) to optimize memory operations, such as extraction, updating, and retrieval, has emerged as a highly promising research direction. However, existing implementations remain highly fragmented and task-specific, lacking a unified infrastructure to streamline the integration, training, and evaluation of these complex pipelines. To address this gap, we present MemFactory, the first unified, highly modular training and inference framework specifically designed for memory-augmented agents. Inspired by the success of unified fine-tuning frameworks like LLaMA-Factory, MemFactory abstracts the memory lifecycle into atomic, plug-and-play components, enabling researchers to seamlessly construct custom memory agents via a "Lego-like" architecture. Furthermore, the framework natively integrates Group Relative Policy Optimization (GRPO) to fine-tune internal memory management policies driven by multi-dimensional environmental rewards. MemFactory provides out-of-the-box support for recent cutting-edge paradigms, including Memory-R1, RMM, and MemAgent. We empirically validate MemFactory on the open-source MemAgent architecture using its publicly available training and evaluation data. Across the evaluation sets, MemFactory improves performance over the corresponding base models on average, with relative gains of up to 14.8%. By providing a standardized, extensible, and easy-to-use infrastructure, MemFactory significantly lowers the barrier to entry, paving the way for future innovations in memory-driven AI agents.

Membership Inference Attacks (MIAs) act as a crucial auditing tool for the opaque training data of Large Language Models (LLMs). However, existing techniques predominantly rely on inaccessible model internals (e.g., logits) or suffer from poor generalization across domains in strict black-box settings where only generated text is available. In this work, we propose SimMIA, a robust MIA framework tailored for this text-only regime by leveraging an advanced sampling strategy and scoring mechanism. Furthermore, we present WikiMIA-25, a new benchmark curated to evaluate MIA performance on modern proprietary LLMs. Experiments demonstrate that SimMIA achieves state-of-the-art results in the black-box setting, rivaling baselines that exploit internal model information.

Singular Value Decomposition (SVD) has recently seen a surge of interest as a simple yet powerful tool for large language models (LLMs) compression, with a growing number of works demonstrating 20-80% parameter reductions at minimal accuracy loss. Previous SVD-based approaches have focused primarily on reducing the memory footprint of model weights, largely overlooking the additional activation memory overhead incurred during inference when applying truncated factors via standard dense CUDA kernels. Our experiments demonstrate that this activation overhead, scaling with sequence length and hidden dimension, prevents current SVD compression techniques from achieving any reduction in peak inference memory, thereby limiting their viability for real-world, on-device deployments. We introduce FlashSVD, a novel, end-to-end rank-aware streaming inference framework specifically designed for SVD-compressed large language models. FlashSVD can be seamlessly integrated with any model that employs SVD-based methods for parameter reduction. By fusing low-rank projection kernels directly into both the self-attention and feed-forward network (FFN) pipelines, FlashSVD avoid materializing full-size activation buffers. Instead, small tiles of the truncated factors are loaded into on-chip SRAM, multiplied and reduced on the fly, and immediately evicted, preserving high GPU occupancy and adding no extra latency. On standard encoder benchmarks (e.g., BERT-Base), FlashSVD cuts peak activation memory by up to 70.2% and intermediate transient memory by 75%, all while incur no accuracy loss with upstreaming compression methods, offering a practical path toward memory-constrained deployment of low-rank LLMs.

While Large Language Models (LLMs) have achieved remarkable performance, they remain vulnerable to jailbreak attacks that circumvent safety constraints. Existing strategies, ranging from heuristic prompt engineering to computationally intensive optimization, often face significant trade-offs between effectiveness and efficiency. In this work, we propose Contextual Representation Ablation (CRA), a novel inference-time intervention framework designed to dynamically silence model guardrails. Predicated on the geometric insight that refusal behaviors are mediated by specific low-rank subspaces within the model's hidden states, CRA identifies and suppresses these refusal-inducing activation patterns during decoding without requiring expensive parameter updates or training. Empirical evaluation across multiple safety-aligned open-source LLMs demonstrates that CRA significantly outperforms baselines. These results expose the intrinsic fragility of current alignment mechanisms, revealing that safety constraints can be surgically ablated from internal representations, and underscore the urgent need for more robust defenses that secure the model's latent space.

The growing demand for text-to-image generation has led to rapid advances in generative modeling. Recently, text-to-image diffusion models trained with flow matching algorithms, such as FLUX, have achieved remarkable progress and emerged as strong alternatives to conventional diffusion models. At the same time, inference-time enhancement strategies have been shown to improve the generation quality and text-prompt alignment of text-to-image diffusion models. However, these techniques are mainly applicable to conventional diffusion models and usually fail to perform well on flow models. To bridge this gap, we propose Reflective Flow Sampling (RF-Sampling), a theoretically-grounded and training-free inference enhancement framework explicitly designed for flow models, especially for the CFG-distilled variants (i.e., models distilled from CFG guidance techniques), like FLUX. Departing from heuristic interpretations, we provide a formal derivation proving that RF-Sampling implicitly performs gradient ascent on the text-image alignment score. By leveraging a linear combination of textual representations and integrating them with flow inversion, RF-Sampling allows the model to explore noise spaces that are more consistent with the input prompt. Extensive experiments across multiple benchmarks demonstrate that RF-Sampling consistently improves both generation quality and prompt alignment. Moreover, RF-Sampling is also the first inference enhancement method that can exhibit test-time scaling ability to some extent on FLUX.

Diffusion-based vision-language-action models (dVLAs) are promising for embodied intelligence but are fundamentally limited in real-time deployment by the high latency of full inference. We propose Realtime-VLA FLASH, a speculative inference framework that eliminates most full inference calls during replanning by introducing a lightweight draft model with parallel verification via the main model's Action Expert and a phase-aware fallback mechanism that reverts to the full inference pipeline when needed. This design enables low-latency, high-frequency replanning without sacrificing reliability. Experiments show that on LIBERO, FLASH largely preserves task performance by replacing many 58.0 ms full-inference rounds with speculative rounds as fast as 7.8 ms, lowering task-level average inference latency to 19.1 ms (3.04x speedup). We additionally demonstrate effectiveness on real-world conveyor-belt sorting, highlighting its practical impact for latency-critical embodied tasks.

Disaggregated inference has become an essential framework that separates the prefill (P) and decode (D) stages in large language model inference to improve throughput. However, the KV cache transfer faces significant delays between prefill and decode nodes. The block-wise calling method and discontinuous KV cache memory allocation increase the number of calls to the transmission kernel. Additionally, existing frameworks often fix the roles of P and D nodes, leading to computational imbalances. In this paper, we propose FlowKV, a novel disaggregated inference framework, which reduces the average transmission latency of KV cache by 96%, from 0.944s to 0.053s, almost eliminating the transfer time relative to the total request latency by optimizing the KV cache transfer. FlowKV introduces the Load-Aware Scheduler for balanced request scheduling and flexible PD node allocation. This design maximizes hardware resource utilization, achieving peak system throughput across various scenarios, including normal, computational imbalance, and extreme overload conditions. Experimental results demonstrate that FlowKV significantly accelerates inference by 15.2%-48.9% on LongBench dataset compared to the baseline and supports applications with heterogeneous GPUs.

Partial differential equations (PDEs) are fundamental to modeling physical systems, yet solving them remains a complex challenge. Traditional numerical solvers rely on expert knowledge to implement and are computationally expensive, while neural-network-based solvers require large training datasets and often lack interpretability. In this work, we frame PDE solving as a code generation task and introduce CodePDE, the first inference framework for generating PDE solvers using large language models (LLMs). Leveraging advanced inference-time algorithms and scaling strategies, CodePDE unlocks critical capacities of LLM for PDE solving: reasoning, debugging, selfrefinement, and test-time scaling -- all without task-specific tuning. CodePDE achieves superhuman performance across a range of representative PDE problems. We also present a systematic empirical analysis of LLM generated solvers, analyzing their accuracy, efficiency, and numerical scheme choices. Our findings highlight the promise and the current limitations of LLMs in PDE solving, offering a new perspective on solver design and opportunities for future model development. Our code is available at https://github.com/LithiumDA/CodePDE.

Speculative decoding and quantization effectively accelerate memory-bound inference of large language models. Speculative decoding mitigates the memory bandwidth bottleneck by verifying multiple tokens within a single forward pass, which increases computational effort. Quantization achieves this optimization by compressing weights and activations into lower bit-widths and also reduces computations via low-bit matrix multiplications. To further leverage their strengths, we investigate the integration of these two techniques. Surprisingly, experiments applying the advanced speculative decoding method EAGLE-2 to various quantized models reveal that the memory benefits from 4-bit weight quantization are diminished by the computational load from speculative decoding. Specifically, verifying a tree-style draft incurs significantly more time overhead than a single-token forward pass on 4-bit weight quantized models. This finding led to our new speculative decoding design: a hierarchical framework that employs a small model as an intermediate stage to turn tree-style drafts into sequence drafts, leveraging the memory access benefits of the target quantized model. Experimental results show that our hierarchical approach achieves a 2.78times speedup across various tasks for the 4-bit weight Llama-3-70B model on an A100 GPU, outperforming EAGLE-2 by 1.31times. Code available at https://github.com/AI9Stars/SpecMQuant.

The rapid advancement of ML models in critical sectors such as healthcare, finance, and security has intensified the need for robust data security, model integrity, and reliable outputs. Large multimodal foundational models, while crucial for complex tasks, present challenges in scalability, reliability, and potential misuse. Decentralized systems offer a solution by distributing workload and mitigating central points of failure, but they introduce risks of unauthorized access to sensitive data across nodes. We address these challenges with a comprehensive framework designed for responsible AI development. Our approach incorporates: 1) Zero-knowledge proofs for secure model verification, enhancing trust without compromising privacy. 2) Consensus-based verification checks to ensure consistent outputs across nodes, mitigating hallucinations and maintaining model integrity. 3) Split Learning techniques that segment models across different nodes, preserving data privacy by preventing full data access at any point. 4) Hardware-based security through trusted execution environments (TEEs) to protect data and computations. This framework aims to enhance security and privacy and improve the reliability and fairness of multimodal AI systems. Promoting efficient resource utilization contributes to more sustainable AI development. Our state-of-the-art proofs and principles demonstrate the framework's effectiveness in responsibly democratizing artificial intelligence, offering a promising approach for building secure and private foundational models.

Automatic workflow generation is the process of automatically synthesizing sequences of LLM calls, tool invocations, and post-processing steps for complex end-to-end tasks. Most prior methods cast this task as an optimization problem with limited theoretical grounding. We propose to cast workflow generation as Bayesian inference over a posterior distribution on workflows, and introduce Bayesian Workflow Generation (BWG), a sampling framework that builds workflows step-by-step using parallel look-ahead rollouts for importance weighting and a sequential in-loop refiner for pool-wide improvements. We prove that, without the refiner, the weighted empirical distribution converges to the target posterior. We instantiate BWG as BayesFlow, a training-free algorithm for workflow construction. Across six benchmark datasets, BayesFlow improves accuracy by up to 9 percentage points over SOTA workflow generation baselines and by up to 65 percentage points over zero-shot prompting, establishing BWG as a principled upgrade to search-based workflow design. Code will be available on https://github.com/BoYuanVisionary/BayesFlow.

The rapid expansion of distributed Artificial Intelligence (AI) workloads beyond centralized data centers creates a demand for new communication substrates. These substrates must operate reliably in heterogeneous and permissionless environments, where Network Address Translators (NATs) and firewalls impose significant constraints. Existing solutions, however, are either designed for controlled data center deployments or implemented as monolithic systems that tightly couple machine learning logic with networking code. To address these limitations, we present Lattica, a decentralized cross-NAT communication framework designed to support distributed AI systems. Lattica integrates three core components. First, it employs a robust suite of NAT traversal mechanisms to establish a globally addressable peer-to-peer mesh. Second, it provides a decentralized data store based on Conflict-free Replicated Data Types (CRDTs), ensuring verifiable and eventually consistent state replication. Third, it incorporates a content discovery layer that leverages distributed hash tables (DHTs) together with an optimized RPC protocol for efficient model synchronization. By integrating these components, Lattica delivers a complete protocol stack for sovereign, resilient, and scalable AI systems that operate independently of centralized intermediaries. It is directly applicable to edge intelligence, collaborative reinforcement learning, and other large-scale distributed machine learning scenarios.

Large Language Models (LLMs) have demonstrated remarkable fluency and versatility across a wide range of NLP tasks, yet they remain prone to factual inaccuracies and hallucinations. This limitation poses significant risks in high-stakes domains such as healthcare, law, and scientific communication, where trust and verifiability are paramount. In this paper, we introduce DAVinCI - a Dual Attribution and Verification framework designed to enhance the factual reliability and interpretability of LLM outputs. DAVinCI operates in two stages: (i) it attributes generated claims to internal model components and external sources; (ii) it verifies each claim using entailment-based reasoning and confidence calibration. We evaluate DAVinCI across multiple datasets, including FEVER and CLIMATE-FEVER, and compare its performance against standard verification-only baselines. Our results show that DAVinCI significantly improves classification accuracy, attribution precision, recall, and F1-score by 5-20%. Through an extensive ablation study, we isolate the contributions of evidence span selection, recalibration thresholds, and retrieval quality. We also release a modular DAVinCI implementation that can be integrated into existing LLM pipelines. By bridging attribution and verification, DAVinCI offers a scalable path to auditable, trustworthy AI systems. This work contributes to the growing effort to make LLMs not only powerful but also accountable.

With ChatGPT as a representative, tons of companies have began to provide services based on large Transformers models. However, using such a service inevitably leak users' prompts to the model provider. Previous studies have studied secure inference for Transformer models using secure multiparty computation (MPC), where model parameters and clients' prompts are kept secret. Despite this, these frameworks are still limited in terms of model performance, efficiency, and deployment. To address these limitations, we propose framework PUMA to enable fast and secure Transformer model inference. Our framework designs high quality approximations for expensive functions, such as GeLU and Softmax, which significantly reduce the cost of secure inference while preserving the model performance. Additionally, we design secure Embedding and LayerNorm procedures that faithfully implement the desired functionality without undermining the Transformer architecture. PUMA is about 2x faster than the state-of-the-art MPC framework MPCFORMER(ICLR 2023) and has similar accuracy as plaintext models without fine-tuning (which the previous works failed to achieve). One more thing, PUMA can evaluate LLaMA-7B in around 5 minutes to generate 1 token. To our best knowledge, this is the first time that a model with such a parameter size is able to be evaluated under MPC. PUMA has been open-sourced in the Github repository of SecretFlow-SPU.

The rapid growth of large-scale AI models, particularly large language models has brought significant challenges in data privacy, computational resources, and accessibility. Traditional centralized architectures often struggle to meet required data security and scalability needs which hinders the democratization of AI systems. Nesa introduces a model-agnostic sharding framework designed for decentralized AI inference. Our framework uses blockchain-based sequential deep neural network sharding to distribute computational tasks across a diverse network of nodes based on a personalised heuristic and routing mechanism. This enables efficient distributed training and inference for recent large-scale models even on consumer-grade hardware. We use compression techniques like dynamic blockwise quantization and mixed matrix decomposition to reduce data transfer and memory needs. We also integrate robust security measures, including hardware-based trusted execution environments to ensure data integrity and confidentiality. Evaluating our system across various natural language processing and vision tasks shows that these compression strategies do not compromise model accuracy. Our results highlight the potential to democratize access to cutting-edge AI technologies by enabling secure and efficient inference on a decentralized network.

Large language models (LLMs) have achieved near-human performance across diverse reasoning tasks, yet their deployment on resource-constrained Internet-of-Things (IoT) devices remains impractical due to massive parameter footprints and memory-intensive autoregressive decoding. While split computing offers a promising solution by partitioning model execution between edge devices and cloud servers, existing approaches fail to address the unique challenges of autoregressive inference, particularly the iterative token generation process and expanding key-value (KV) cache requirements. This work introduces the first autoregressive-aware split computing framework designed explicitly for LLM deployment on edge devices. Our approach makes three key contributions. First, we develop one-point split compression (OPSC), a mixed-precision quantization scheme that prevents out-of-memory failures by strategically partitioning models into front-end and back-end segments with different precision levels. Second, we propose a two-stage intermediate compression pipeline that combines threshold splitting (TS) and token-wise adaptive bit quantization (TAB-Q) to preserve accuracy-critical activations while dramatically reducing communication overhead. Third, we formulate a unified optimization framework that jointly selects optimal split points, quantization settings, and sequence lengths to satisfy strict memory and latency constraints. Extensive evaluations across diverse LLMs and hardware platforms demonstrate superior performance compared to state-of-the-art quantization methods, including SmoothQuant, OmniQuant, and Atom. The framework achieves a 1.49 inference speedup and significant communication overhead reduction while maintaining or improving model accuracy.

The current paradigm for safety alignment of large language models (LLMs) follows a one-size-fits-all approach: the model refuses to interact with any content deemed unsafe by the model provider. This approach lacks flexibility in the face of varying social norms across cultures and regions. In addition, users may have diverse safety needs, making a model with static safety standards too restrictive to be useful, as well as too costly to be re-aligned. We propose Controllable Safety Alignment (CoSA), a framework designed to adapt models to diverse safety requirements without re-training. Instead of aligning a fixed model, we align models to follow safety configs -- free-form natural language descriptions of the desired safety behaviors -- that are provided as part of the system prompt. To adjust model safety behavior, authorized users only need to modify such safety configs at inference time. To enable that, we propose CoSAlign, a data-centric method for aligning LLMs to easily adapt to diverse safety configs. Furthermore, we devise a novel controllability evaluation protocol that considers both helpfulness and configured safety, summarizing them into CoSA-Score, and construct CoSApien, a human-authored benchmark that consists of real-world LLM use cases with diverse safety requirements and corresponding evaluation prompts. We show that CoSAlign leads to substantial gains of controllability over strong baselines including in-context alignment. Our framework encourages better representation and adaptation to pluralistic human values in LLMs, and thereby increasing their practicality.

This paper introduces PowerInfer-2, a framework designed for high-speed inference of Large Language Models (LLMs) on smartphones, particularly effective for models whose sizes exceed the device's memory capacity. The key insight of PowerInfer-2 is to utilize the heterogeneous computation, memory, and I/O resources in smartphones by decomposing traditional matrix computations into fine-grained neuron cluster computations. Specifically, PowerInfer-2 features a polymorphic neuron engine that adapts computational strategies for various stages of LLM inference. Additionally, it introduces segmented neuron caching and fine-grained neuron-cluster-level pipelining, which effectively minimize and conceal the overhead caused by I/O operations. The implementation and evaluation of PowerInfer-2 demonstrate its capability to support a wide array of LLM models on two smartphones, achieving up to a 29.2x speed increase compared with state-of-the-art frameworks. Notably, PowerInfer-2 is the first system to serve the TurboSparse-Mixtral-47B model with a generation rate of 11.68 tokens per second on a smartphone. For models that fit entirely within the memory, PowerInfer-2 can achieve approximately a 40% reduction in memory usage while maintaining inference speeds comparable to llama.cpp and MLC-LLM. For more details, including a demonstration video, please visit the project site at www.powerinfer.ai/v2.

Large language models (LLMs) are increasingly utilized for complex tasks requiring longer context lengths, with some models supporting up to 128K or 1M tokens. This trend, however, presents significant challenges in inference speed and memory management. Quantization emerges as a promising approach to address the widening gap between LLM size and memory capacity. However, traditional quantization schemes often yield suboptimal compression results for KV caches due to two key factors: i) On-the-fly quantization and de-quantization, causing significant performance overhead; ii) Prevalence of outliers in KV values, challenging low-bitwidth uniform quantization. To this end, we propose MILLION, a novel quantization framework achieving low-bitwidth KV cache through product quantization. First, we conduct a thorough analysis of KV cache distribution, revealing the limitations of existing quantization schemes. Second, we introduce a non-uniform quantization algorithm based on product quantization, which efficiently compresses data while preserving accuracy. Third, we develop a high-performance GPU inference framework with efficient attention kernel and pipeline design for MILLION that leverages sparse computation and asynchronous quantization, significantly enhancing inference speed. Comprehensive evaluation results demonstrate that MILLION can achieve 4 bits quantization with trivial perplexity and accuracy loss, and achieve 2.09x end-to-end performance gains at 32K context length. Code is released at https://github.com/ZongwuWang/MILLION.

This paper introduces KunLunBaizeRAG, a reinforcement learning-driven reasoning framework designed to enhance the reasoning capabilities of large language models (LLMs) in complex multi-hop question-answering tasks. The framework addresses key limitations of traditional RAG, such as retrieval drift, information redundancy, and strategy rigidity. Key innovations include the RAG-driven Reasoning Alignment (RDRA) mechanism, the Search-Think Iterative Enhancement (STIE) mechanism, the Network-Local Intelligent Routing (NLR) mechanism, and a progressive hybrid training strategy. Experimental results demonstrate significant improvements in exact match (EM) and LLM-judged score (LJ) across four benchmarks, highlighting the framework's robustness and effectiveness in complex reasoning scenarios.

Recent advances in large language models (LLMs) have accelerated progress toward artificial general intelligence, with inference-time scaling emerging as a key technique. Contemporary approaches leverage either sequential reasoning (iteratively extending chains of thought) or parallel reasoning (generating multiple solutions simultaneously) to scale inference. However, both paradigms face fundamental limitations: sequential scaling typically relies on arbitrary token budgets for termination, leading to inefficiency or premature cutoff; while parallel scaling often lacks coordination among parallel branches and requires intrusive fine-tuning to perform effectively. In light of these challenges, we aim to design a flexible test-time collaborative inference framework that exploits the complementary strengths of both sequential and parallel reasoning paradigms. Towards this goal, the core challenge lies in developing an efficient and accurate intrinsic quality metric to assess model responses during collaborative inference, enabling dynamic control and early termination of the reasoning trace. To address this challenge, we introduce semantic entropy (SE), which quantifies the semantic diversity of parallel model responses and serves as a robust indicator of reasoning quality due to its strong negative correlation with accuracy...

Recent advances in leveraging the agentic paradigm of large language models (LLMs) utilization have significantly enhanced Text-to-SQL capabilities, enabling users without specialized database expertise to query data intuitively. However, deploying these agentic LLM-based Text-to-SQL systems in production poses substantial challenges due to their inherently multi-stage workflows, stringent latency constraints, and potentially heterogeneous GPU infrastructure in enterprise environments. Current LLM serving frameworks lack effective mechanisms for handling interdependent inference tasks, dynamic latency variability, and resource heterogeneity, leading to suboptimal performance and frequent service-level objective (SLO) violations. In this paper, we introduce HEXGEN-TEXT2SQL, a novel framework designed explicitly to schedule and execute agentic multi-stage LLM-based Text-to-SQL workflows on heterogeneous GPU clusters that handle multi-tenant end-to-end queries. HEXGEN-TEXT2SQL introduce a hierarchical scheduling approach combining global workload-balanced task dispatching and local adaptive urgency-guided prioritization, guided by a systematic analysis of agentic Text-to-SQL workflows. Additionally, we propose a lightweight simulation-based method for tuning critical scheduling hyperparameters, further enhancing robustness and adaptability. Our extensive evaluation on realistic Text-to-SQL benchmarks demonstrates that HEXGEN-TEXT2SQL significantly outperforms state-of-the-art LLM serving frameworks. Specifically, HEXGEN-TEXT2SQL reduces latency deadlines by up to 1.67times (average: 1.41times) and improves system throughput by up to 1.75times (average: 1.65times) compared to vLLM under diverse, realistic workload conditions. Our code is available at https://github.com/Relaxed-System-Lab/Hexgen-Flow.

Fine-grained Mixture-of-Experts (MoE) models sparsely activate only a subset of experts per token, reducing activated computation while maintaining high model capacity. However, in memory-constrained inference scenarios, only a small set of experts can be cached. Experts not in the cache must be fetched from slow external storage (e.g., UFS), leading to frequent evictions and substantial I/O overhead. We propose ReMoE, a router fine-tuning framework designed to boost token-wise expert reuse. ReMoE biases the router toward recently selected experts, producing temporally stable routing that better matches cache locality constraints. By increasing short-horizon expert reuse, ReMoE reduces expert fetches from storage without adding inference-time computation. Experiments on DeepSeek and Qwen models show that ReMoE improves expert reuse by 26% while maintaining downstream task performance. Real-system evaluations further confirm these benefits, improving output throughput by 8.4% under vLLM GPU-CPU expert offloading and reducing TPOT by 43.6-49.8% under llama.cpp on Jetson Orin NX, corresponding to a 1.77-1.99times decode speedup across diverse workloads. Checkpoints and usage instructions are available at https://github.com/BUAA-OSCAR/ReMoE.

Vision-Language-Action (VLA) models have achieved remarkable breakthroughs in robotics, with the action chunk playing a dominant role in these advances. Given the real-time and continuous nature of robotic motion control, the strategies for fusing a queue of successive action chunks have a profound impact on the overall performance of VLA models. Existing methods suffer from jitter, stalling, or even pauses in robotic action execution, which not only limits the achievable execution speed but also reduces the overall success rate of task completion. This paper introduces VLA-RAIL (A Real-Time Asynchronous Inference Linker), a novel framework designed to address these issues by conducting model inference and robot motion control asynchronously and guaranteeing smooth, continuous, and high-speed action execution. The core contributions of the paper are two fold: a Trajectory Smoother that effectively filters out the noise and jitter in the trajectory of one action chunk using polynomial fitting and a Chunk Fuser that seamlessly align the current executing trajectory and the newly arrived chunk, ensuring position, velocity, and acceleration continuity between two successive action chunks. We validate the effectiveness of VLA-RAIL on a benchmark of dynamic simulation tasks and several real-world manipulation tasks. Experimental results demonstrate that VLA-RAIL significantly reduces motion jitter, enhances execution speed, and improves task success rates, which will become a key infrastructure for the large-scale deployment of VLA models.

We introduce AIBrix, a cloud-native, open-source framework designed to optimize and simplify large-scale LLM deployment in cloud environments. Unlike traditional cloud-native stacks, AIBrix follows a co-design philosophy, ensuring every layer of the infrastructure is purpose-built for seamless integration with inference engines like vLLM. AIBrix introduces several key innovations to reduce inference costs and enhance performance including high-density LoRA management for dynamic adapter scheduling, LLM-specific autoscalers, and prefix-aware, load-aware routing. To further improve efficiency, AIBrix incorporates a distributed KV cache, boosting token reuse across nodes, leading to a 50% increase in throughput and a 70% reduction in inference latency. AIBrix also supports unified AI runtime which streamlines model management while maintaining vendor-agnostic engine compatibility. For large-scale multi-node inference, AIBrix employs hybrid orchestration -- leveraging Kubernetes for coarse-grained scheduling and Ray for fine-grained execution -- to balance efficiency and flexibility. Additionally, an SLO-driven GPU optimizer dynamically adjusts resource allocations, optimizing heterogeneous serving to maximize cost efficiency while maintaining service guarantees. Finally, AIBrix enhances system reliability with AI accelerator diagnostic tools, enabling automated failure detection and mock-up testing to improve fault resilience. AIBrix is available at https://github.com/vllm-project/aibrix.

In single-cell research, tracing and analyzing high-throughput single-cell differentiation trajectories is crucial for understanding biological processes. Key to this is the robust modeling of hierarchical structures that govern cellular development. Traditional methods face limitations in computational cost, performance, and stability. VAE-based approaches have made strides but still require branch-specific network modules, limiting their scalability and stability, while often suffering from posterior collapse. To overcome these challenges, we introduce HDTree, a generative modeling framework designed for robust lineage inference. HDTree captures tree relationships within a hierarchical latent space using a unified hierarchical codebook and employs a quantized diffusion process to model continuous cell state transitions. By aligning the generative process with the Waddington landscape, this method not only improves stability and scalability but also enhances the biological plausibility of inferred lineages. HDTree's effectiveness is demonstrated through comparisons on both general-purpose and single-cell datasets, where it outperforms existing methods in lineage inference accuracy, reconstruction quality, and hierarchical consistency. These contributions enable accurate and efficient modeling of cellular differentiation paths, offering reliable insights for biological discovery.\footnote{Code is available at https://github.com/zangzelin/code\_HDTree\_icml.

Despite the computational efficiency of MoE models, the excessive memory footprint and I/O overhead inherent in multi-expert architectures pose formidable challenges for real-time inference on resource-constrained edge platforms. While existing static methods struggle with a rigid latency-accuracy trade-off, we observe that expert importance is highly skewed and depth-dependent. Motivated by these insights, we propose DyMoE, a dynamic mixed-precision quantization framework designed for high-performance edge inference. Leveraging insights into expert importance skewness and depth-dependent sensitivity, DyMoE introduces: (1) importance-aware prioritization to dynamically quantize experts at runtime; (2) depth-adaptive scheduling to preserve semantic integrity in critical layers; and (3) look-ahead prefetching to overlap I/O stalls. Experimental results on commercial edge hardware show that DyMoE reduces Time-to-First-Token (TTFT) by 3.44x-22.7x and up to a 14.58x speedup in Time-Per-Output-Token (TPOT) compared to state-of-the-art offloading baselines, enabling real-time, accuracy-preserving MoE inference on resource-constrained edge devices.

Deep research agents autonomously conduct open-ended investigations, integrating complex information retrieval with multi-step reasoning across diverse sources to solve real-world problems. To sustain this capability on long-horizon tasks, reliable verification is critical during both training and inference. A major bottleneck in existing paradigms stems from the lack of explicit verification mechanisms in QA data synthesis, trajectory construction, and test-time scaling. Errors introduced at each stage propagate downstream and degrade the overall agent performance. To address this, we present Marco DeepResearch, a deep research agent optimized with a verification-centric framework design at three levels: (1)~QA Data Synthesis: We introduce verification mechanisms to graph-based and agent-based QA synthesis to control question difficulty while ensuring answers are unique and correct; (2)~Trajectory Construction: We design a verification-driven trajectory synthesis method that injects explicit verification patterns into training trajectories; and (3)~Test-time scaling: We use Marco DeepResearch itself as a verifier at inference time and effectively improve performance on challenging questions. Extensive experimental results demonstrate that our proposed Marco DeepResearch agent significantly outperforms 8B-scale deep research agents on most challenging benchmarks, such as BrowseComp and BrowseComp-ZH. Crucially, under a maximum budget of 600 tool calls, Marco DeepResearch even surpasses or approaches several 30B-scale agents, like Tongyi DeepResearch-30B.

Speculative Decoding (SD) has emerged as a premier technique for accelerating Large Language Model (LLM) inference by decoupling token generation into rapid drafting and parallel verification. While recent advancements in self-speculation and lookahead decoding have successfully minimized drafting overhead, they have shifted the primary performance bottleneck to the verification phase. Since verification requires a full forward pass of the target model, it remains strictly memory-bandwidth bound, fundamentally limiting the maximum achievable speedup.In this paper, we introduce Quasar (Quantized Self-speculative Acceleration for Rapid Inference), a novel, training-free framework designed to overcome this "memory wall" by employing low-bit quantization specifically for the verification stage. Our empirical analysis reveals that while aggressive structural pruning significantly degrades verification accuracy, quantization-based verification preserves the logit distribution with high fidelity while effectively halving memory traffic. Extensive experiments on state-of-the-art models (e.g., OpenPangu and Qwen3) demonstrate that Quasar maintains a speculative acceptance length comparable to full-precision methods while achieving a 1.28times improvement in end-to-end throughput. Being orthogonal to existing drafting strategies, Quasar offers a generic and efficient pathway to accelerate the verification leg of speculative execution. Code is available at https://github.com/Tom-HG/Quasar.

As an increasing majority of global video content is consumed on social platforms for interactive social purposes, video generation models built for social worlds are important but largely overlooked by previous studies. In this work, we define the position of social world models and build a prototype model as the first step towards this goal. While previous world models successfully simulate physical environments or gaming world exploration, they remain fundamentally detached from human-centric social dynamics. To bridge this gap as the first step to social world models, we present MaineCoon, the first real-time audio-visual autoregressive model that has 22B parameters and is capable of real-time streaming generation and sub-second interaction, with a record-breaking frame rate of up to 47.5 FPS, on a single GPU. To the best of our knowledge, MaineCoon is also the first real-time audio-visual generation model specifically optimized for social-interactive applications. To enable efficient and stable training, we introduce several novel techniques into MaineCoon, including self-resampling, cross-modal representation alignment, domain-aware preference optimization, and reinforced online-policy distillation (ROPD). We also design the first agentic streaming inference framework that supports thousand-second-scale or even longer generation while mitigating drift with agentic cache management and prompt planing. These innovations significantly accelerate training while optimizing real-time inference performance. We believe this work not only sets a new state-of-the-art (SOTA) performance benchmark for high-quality, low-latency, and long-horizon audio-visual autoregressive models, but also points out the paradigm shift desired for next-generation AI-native social platforms.

Deploying large language models (LLMs) on mobile devices is an emerging trend to enable data privacy and offline accessibility of LLM applications. Modern mobile neural processing units (NPUs) make such deployment increasingly feasible. However, existing mobile LLM inference frameworks suffer from high start-up latency due to their inevitable cold starts, i.e., launching LLM inferences when the model is not hosted in device memory. In this paper, we identify the key bottleneck of mobile LLM cold starts as the waste of flash bandwidth on unimportant model parameters. We design EdgeFlow, a mobile LLM inference framework that mitigates the cold start issue by adaptively adjusting the precisions of LLM parameters. Specifically, EdgeFlow leverages 1) an NPU-aware adaptive quantization algorithm that assigns different precisions to weights in a finer granularity according to their importance and NPU constraints, 2) an SIMD-friendly packing format that accelerates the transformation of various-precision weights into fixed-sized NPU-native data types, and 3) a synergistic granular pipeline that coordinates CPU and NPU computation in a fine-grained and dynamic manner. Experimental results show that EdgeFlow reduces cold-start latency by up to 4.07x compared with three state-of-the-art mobile LLM inference frameworks, i.e., llama.cpp, MNN, and llm.npu, under comparable model accuracy.

As global e-commerce rapidly expands into emerging markets, the lack of high-quality semantic representations for low-resource languages has become a decisive bottleneck for retrieval, recommendation, and search systems. In this work, we present Compass-Embedding v4, a high-efficiency multilingual embedding framework specifically optimized for Southeast Asian (SEA) e-commerce scenarios, where data scarcity, noisy supervision, and strict production constraints jointly challenge representation learning. Compass-Embedding v4 addresses three core challenges. First, large-batch contrastive training under mixed task supervision introduces systematic false negatives that degrade semantic alignment. We propose Class-Aware Masking (CAM), a lightweight modification to the InfoNCE objective that suppresses invalid in-batch negatives and improves semantic discrimination without altering training efficiency. Second, low-resource SEA languages suffer from limited and uneven data coverage. We construct a diversified training corpus through context-grounded synthetic data generation, cross-lingual translation, and structured e-commerce data construction, enabling robust multilingual and domain-specific learning. Third, production deployment requires high-throughput inference while preserving embedding quality. We combine robustness-driven large-batch training with spherical model merging to mitigate catastrophic forgetting, and optimize inference via vLLM and FP8 quantization. Extensive evaluations across multilingual benchmarks and proprietary e-commerce tasks show that Compass-Embedding v4 achieves state-of-the-art performance on major SEA languages, significantly outperforming general-purpose embedding models in domain-specific retrieval and classification, while maintaining competitive performance on high-resource languages.

Chain-of-thought (CoT) reasoning has been highly successful in solving complex tasks in natural language processing, and recent multimodal large language models (MLLMs) have extended this paradigm to video reasoning. However, these models typically build on lengthy reasoning chains and large numbers of input visual tokens. Motivated by empirical observations from our benchmark study, we hypothesize that concise reasoning combined with a reduced set of visual tokens can be sufficient for effective video reasoning. To evaluate this hypothesis, we design and validate an efficient post-training and inference framework that enhances a video MLLM's reasoning capability. Our framework enables models to operate on compressed visual tokens and generate brief reasoning traces prior to answering. The resulting models achieve substantially improved inference efficiency, deliver competitive performance across diverse benchmarks, and avoid reliance on manual CoT annotations or supervised fine-tuning. Collectively, our results suggest that long, human-like CoT reasoning may not be necessary for general video reasoning, and that concise reasoning can be both effective and efficient. Our code will be released at https://github.com/LaVi-Lab/Rethink_CoT_Video.

Cross-view geo-localization (CVGL) plays a vital role in drone-based multimedia applications, enabling precise localization by matching drone-captured aerial images against geo-tagged satellite databases in GNSS-denied environments. However, existing methods rely on resource-intensive feature alignment and multi-branch architectures, incurring high inference costs that limit their deployment on edge devices. We propose MobileGeo, a mobile-friendly framework designed for efficient on-device CVGL: 1) During training, a Hierarchical Distillation (HD-CVGL) paradigm, coupled with Uncertainty-Aware Prediction Alignment (UAPA), distills essential information into a compact model without incurring inference overhead. 2) During inference, an efficient Multi-view Selection Refinement Module (MSRM) leverages mutual information to filter redundant views and reduce computational load. Extensive experiments demonstrate that MobileGeo outperforms previous state-of-the-art methods, achieving a 4.19% improvement in AP on University1652 dataset while being over 5 times efficient in FLOPs and 3 times faster. Crucially, MobileGeo runs at 251.5 FPS on an NVIDIA AGX Orin edge device, demonstrating its practical viability for real-time on-device drone geo-localization. The code is available at https://github.com/SkyEyeLoc/MobileGeo.

Vision-Language-Action (VLA) models have emerged as a powerful paradigm for open-world robot manipulation, but their practical deployment is often constrained by cost: billion-scale VLM backbones and iterative diffusion/flow-based action heads incur high latency and compute, making real-time control expensive on commodity hardware. We present A1, a fully open-source and transparent VLA framework designed for low-cost, high-throughput inference without sacrificing manipulation success; Our approach leverages pretrained VLMs that provide implicit affordance priors for action generation. We release the full training stack (training code, data/data-processing pipeline, intermediate checkpoints, and evaluation scripts) to enable end-to-end reproducibility. Beyond optimizing the VLM alone, A1 targets the full inference pipeline by introducing a budget-aware adaptive inference scheme that jointly accelerates the backbone and the action head. Specifically, we monitor action consistency across intermediate VLM layers to trigger early termination, and propose Inter-Layer Truncated Flow Matching that warm-starts denoising across layers, enabling accurate actions with substantially fewer effective denoising iterations. Across simulation benchmarks (LIBERO, VLABench) and real robots (Franka, AgiBot), A1 achieves state-of-the-art success rates while significantly reducing inference cost (e.g., up to 72% lower per-episode latency for flow-matching inference and up to 76.6% backbone computation reduction with minor performance degradation). On RoboChallenge, A1 achieves an average success rate of 29.00%, outperforming baselines including pi0(28.33%), X-VLA (21.33%), and RDT-1B (15.00%).

As the range of applications for Large Language Models (LLMs) continues to grow, the demand for effective serving solutions becomes increasingly critical. Despite the versatility of LLMs, no single model can optimally address all tasks and applications, particularly when balancing performance with cost. This limitation has led to the development of LLM routing systems, which combine the strengths of various models to overcome the constraints of individual LLMs. Yet, the absence of a standardized benchmark for evaluating the performance of LLM routers hinders progress in this area. To bridge this gap, we present RouterBench, a novel evaluation framework designed to systematically assess the efficacy of LLM routing systems, along with a comprehensive dataset comprising over 405k inference outcomes from representative LLMs to support the development of routing strategies. We further propose a theoretical framework for LLM routing, and deliver a comparative analysis of various routing approaches through RouterBench, highlighting their potentials and limitations within our evaluation framework. This work not only formalizes and advances the development of LLM routing systems but also sets a standard for their assessment, paving the way for more accessible and economically viable LLM deployments. The code and data are available at https://github.com/withmartian/routerbench.

Code translation is a crucial task in software development and maintenance. While recent advancements in large language models (LLMs) have improved automated code translation accuracy, these gains often come at the cost of increased inference latency, hindering real-world development workflows that involve human-in-the-loop inspection. To address this trade-off, we propose EffiReasonTrans, a training framework designed to improve translation accuracy while balancing inference latency. We first construct a high-quality reasoning-augmented dataset by prompting a stronger language model, DeepSeek-R1, to generate intermediate reasoning and target translations. Each (source code, reasoning, target code) triplet undergoes automated syntax and functionality checks to ensure reliability. Based on this dataset, we employ a two-stage training strategy: supervised fine-tuning on reasoning-augmented samples, followed by reinforcement learning to further enhance accuracy and balance inference latency. We evaluate EffiReasonTrans on six translation pairs. Experimental results show that it consistently improves translation accuracy (up to +49.2% CA and +27.8% CodeBLEU compared to the base model) while reducing the number of generated tokens (up to -19.3%) and lowering inference latency in most cases (up to -29.0%). Ablation studies further confirm the complementary benefits of the two-stage training framework. Additionally, EffiReasonTrans demonstrates improved translation accuracy when integrated into agent-based frameworks. Our code and data are available at https://github.com/DeepSoftwareAnalytics/EffiReasonTrans.

Aligning with personalized preferences, which vary significantly across cultural, educational, and political differences, poses a significant challenge due to the computational costs and data demands of traditional alignment methods. In response, this paper presents Personalized Alignment at Decoding-time (PAD), a novel framework designed to align LLM outputs with diverse personalized preferences during the inference phase, eliminating the need for additional training. By introducing a unique personalized reward modeling strategy, this framework decouples the text generation process from personalized preferences, facilitating the generation of generalizable token-level personalized rewards. The PAD algorithm leverages these rewards to guide the decoding process, dynamically tailoring the base model's predictions to personalized preferences. Extensive experimental results demonstrate that PAD not only outperforms existing training-based alignment methods in terms of aligning with diverse preferences but also shows significant generalizability to preferences unseen during training and scalability across different base models. This work advances the capability of LLMs to meet user needs in real-time applications, presenting a substantial step forward in personalized LLM alignment.

Recent advances in Video Large Language Models (VLLMs) have significantly enhanced their ability to understand video content. Nonetheless, processing long videos remains challenging due to high computational demands and the redundancy present in the visual data. In this work, we propose METok, a training-free, Multi-stage Event-based Token compression framework designed to accelerate VLLMs' inference while preserving accuracy. METok progressively eliminates redundant visual tokens across three critical stages: (1) event-aware compression during vision encoding, (2) hierarchical token pruning in the prefilling stage based on semantic alignment and event importance, and (3) a decoding-stage KV Cache optimization that further reduces memory consumption. Our experiments on diverse video benchmarks demonstrate that METok achieves an optimal trade-off between efficiency and accuracy by dynamically selecting informative visual tokens. For instance, equipping LongVA-7B with METok realizes an 80.6% FLOPs reduction and 93.5% KV Cache memory savings, all while maintaining comparable or even superior accuracy.

We present Ring-1T, the first open-source, state-of-the-art thinking model with a trillion-scale parameter. It features 1 trillion total parameters and activates approximately 50 billion per token. Training such models at a trillion-parameter scale introduces unprecedented challenges, including train-inference misalignment, inefficiencies in rollout processing, and bottlenecks in the RL system. To address these, we pioneer three interconnected innovations: (1) IcePop stabilizes RL training via token-level discrepancy masking and clipping, resolving instability from training-inference mismatches; (2) C3PO++ improves resource utilization for long rollouts under a token budget by dynamically partitioning them, thereby obtaining high time efficiency; and (3) ASystem, a high-performance RL framework designed to overcome the systemic bottlenecks that impede trillion-parameter model training. Ring-1T delivers breakthrough results across critical benchmarks: 93.4 on AIME-2025, 86.72 on HMMT-2025, 2088 on CodeForces, and 55.94 on ARC-AGI-v1. Notably, it attains a silver medal-level result on the IMO-2025, underscoring its exceptional reasoning capabilities. By releasing the complete 1T parameter MoE model to the community, we provide the research community with direct access to cutting-edge reasoning capabilities. This contribution marks a significant milestone in democratizing large-scale reasoning intelligence and establishes a new baseline for open-source model performance.

To address the challenge of increasing network size, researchers have developed sparse models through network pruning. However, maintaining model accuracy while achieving significant speedups on general computing devices remains an open problem. In this paper, we present a novel mobile inference acceleration framework SparseByteNN, which leverages fine-grained kernel sparsity to achieve real-time execution as well as high accuracy. Our framework consists of two parts: (a) A fine-grained kernel sparsity schema with a sparsity granularity between structured pruning and unstructured pruning. It designs multiple sparse patterns for different operators. Combined with our proposed whole network rearrangement strategy, the schema achieves a high compression rate and high precision at the same time. (b) Inference engine co-optimized with the sparse pattern. The conventional wisdom is that this reduction in theoretical FLOPs does not translate into real-world efficiency gains. We aim to correct this misconception by introducing a family of efficient sparse kernels for ARM and WebAssembly. Equipped with our efficient implementation of sparse primitives, we show that sparse versions of MobileNet-v1 outperform strong dense baselines on the efficiency-accuracy curve. Experimental results on Qualcomm 855 show that for 30% sparse MobileNet-v1, SparseByteNN achieves 1.27x speedup over the dense version and 1.29x speedup over the state-of-the-art sparse inference engine MNN with a slight accuracy drop of 0.224%. The source code of SparseByteNN will be available at https://github.com/lswzjuer/SparseByteNN

High-quality text representations are crucial for natural language understanding (NLU), but low-resource languages like Vietnamese face challenges due to limited annotated data. While pre-trained models like PhoBERT and CafeBERT perform well, their effectiveness is constrained by data scarcity. Contrastive learning (CL) has recently emerged as a promising approach for improving sentence representations, enabling models to effectively distinguish between semantically similar and dissimilar sentences. We propose ViCLSR (Vietnamese Contrastive Learning for Sentence Representations), a novel supervised contrastive learning framework specifically designed to optimize sentence embeddings for Vietnamese, leveraging existing natural language inference (NLI) datasets. Additionally, we propose a process to adapt existing Vietnamese datasets for supervised learning, ensuring compatibility with CL methods. Our experiments demonstrate that ViCLSR significantly outperforms the powerful monolingual pre-trained model PhoBERT on five benchmark NLU datasets such as ViNLI (+6.97% F1), ViWikiFC (+4.97% F1), ViFactCheck (+9.02% F1), UIT-ViCTSD (+5.36% F1), and ViMMRC2.0 (+4.33% Accuracy). ViCLSR shows that supervised contrastive learning can effectively address resource limitations in Vietnamese NLU tasks and improve sentence representation learning for low-resource languages. Furthermore, we conduct an in-depth analysis of the experimental results to uncover the factors contributing to the superior performance of contrastive learning models. ViCLSR is released for research purposes in advancing natural language processing tasks.

Multi-agent systems (MAS) leveraging the impressive capabilities of Large Language Models (LLMs) hold significant potential for tackling complex tasks. However, most current MAS depend on manually designed agent roles and communication protocols. These manual designs often fail to align with the underlying LLMs' strengths and struggle to adapt to novel tasks. Recent automatic MAS approaches attempt to mitigate these limitations but typically necessitate a validation set for tuning and yield static MAS designs lacking adaptability during inference. We introduce MAS-ZERO, the first self-evolved, inference-time framework for automatic MAS design. MAS-ZERO employs meta-level design to iteratively generate, evaluate, and refine MAS configurations tailored to each problem instance, without requiring a validation set. Critically, it enables dynamic agent composition and problem decomposition through meta-feedback on solvability and completeness. Experiments across math, graduate-level QA, and software engineering benchmarks, using both closed-source and open-source LLM backbones of varying sizes, demonstrate that MAS-ZERO outperforms both manual and automatic MAS baselines, achieving a 7.44% average accuracy improvement over the next strongest baseline while maintaining cost-efficiency. These findings underscore the promise of meta-level self-evolved design for creating effective and adaptive MAS.

Lossless model compression holds tremendous promise for alleviating the memory and bandwidth bottlenecks in bit-exact Large Language Model (LLM) serving. However, existing approaches often result in substantial inference slowdowns due to fundamental design mismatches with GPU architectures: at the kernel level, variable-length bitstreams produced by traditional entropy codecs break SIMT parallelism; at the system level, decoupled pipelines lead to redundant memory traffic. We present ZipServ, a lossless compression framework co-designed for efficient LLM inference. ZipServ introduces Tensor-Core-Aware Triple Bitmap Encoding (TCA-TBE), a novel fixed-length format that enables constant-time, parallel decoding, together with a fused decompression-GEMM (ZipGEMM) kernel that decompresses weights on-the-fly directly into Tensor Core registers. This "load-compressed, compute-decompressed" design eliminates intermediate buffers and maximizes compute intensity. Experiments show that ZipServ reduces the model size by up to 30%, achieves up to 2.21x kernel-level speedup over NVIDIA's cuBLAS, and expedites end-to-end inference by an average of 1.22x over vLLM. ZipServ is the first lossless compression system that provides both storage savings and substantial acceleration for LLM inference on GPUs.

Video large language models (VideoLLM) excel at video understanding, but face efficiency challenges due to the quadratic complexity of abundant visual tokens. Our systematic analysis of token compression methods for VideoLLMs reveals two critical issues: (i) overlooking distinctive visual signals across frames, leading to information loss; (ii) suffering from implementation constraints, causing incompatibility with modern architectures or efficient operators. To address these challenges, we distill three design principles for VideoLLM token compression and propose a plug-and-play inference acceleration framework "Video Compression Commander" (VidCom2). By quantifying each frame's uniqueness, VidCom2 adaptively adjusts compression intensity across frames, effectively preserving essential information while reducing redundancy in video sequences. Extensive experiments across various VideoLLMs and benchmarks demonstrate the superior performance and efficiency of our VidCom2. With only 25% visual tokens, VidCom2 achieves 99.6% of the original performance on LLaVA-OV while reducing 70.8% of the LLM generation latency. Notably, our Frame Compression Adjustment strategy is compatible with other token compression methods to further improve their performance. Our code is available at https://github.com/xuyang-liu16/VidCom2.

Mixture-of-Experts (MoE) models are designed to enhance the efficiency of large language models (LLMs) without proportionally increasing the computational demands. However, their deployment on edge devices still faces significant challenges due to high on-demand loading overheads from managing sparsely activated experts. This paper introduces AdapMoE, an algorithm-system co-design framework for efficient MoE inference. AdapMoE features adaptive expert gating and management to reduce the on-demand loading overheads. We observe the heterogeneity of experts loading across layers and tokens, based on which we propose a sensitivity-based strategy to adjust the number of activated experts dynamically. Meanwhile, we also integrate advanced prefetching and cache management techniques to further reduce the loading latency. Through comprehensive evaluations on various platforms, we demonstrate AdapMoE consistently outperforms existing techniques, reducing the average number of activated experts by 25% and achieving a 1.35x speedup without accuracy degradation. Code is available at: https://github.com/PKU-SEC-Lab/AdapMoE.

Multi-task learning (MTL) encapsulates multiple learned tasks in a single model and often lets those tasks learn better jointly. However, when deploying MTL onto those real-world systems that are often resource-constrained or latency-sensitive, two prominent challenges arise: (i) during training, simultaneously optimizing all tasks is often difficult due to gradient conflicts across tasks; (ii) at inference, current MTL regimes have to activate nearly the entire model even to just execute a single task. Yet most real systems demand only one or two tasks at each moment, and switch between tasks as needed: therefore such all tasks activated inference is also highly inefficient and non-scalable. In this paper, we present a model-accelerator co-design framework to enable efficient on-device MTL. Our framework, dubbed M^3ViT, customizes mixture-of-experts (MoE) layers into a vision transformer (ViT) backbone for MTL, and sparsely activates task-specific experts during training. Then at inference with any task of interest, the same design allows for activating only the task-corresponding sparse expert pathway, instead of the full model. Our new model design is further enhanced by hardware-level innovations, in particular, a novel computation reordering scheme tailored for memory-constrained MTL that achieves zero-overhead switching between tasks and can scale to any number of experts. When executing single-task inference, M^{3}ViT achieves higher accuracies than encoder-focused MTL methods, while significantly reducing 88% inference FLOPs. When implemented on a hardware platform of one Xilinx ZCU104 FPGA, our co-design framework reduces the memory requirement by 2.4 times, while achieving energy efficiency up to 9.23 times higher than a comparable FPGA baseline. Code is available at: https://github.com/VITA-Group/M3ViT.

As large language models scale to longer contexts, loading the growing KV cache during attention computation becomes a critical bottleneck. Previous work has shown that attention computation is dominated by a small subset of tokens. This motivates block sparse attention methods that partition the KV cache into fixed-size blocks and selectively compute attention over those blocks exhibiting high importance. However, these methods assign a uniform block size across all attention heads, implicitly assuming homogeneous behavior throughout the model. Our analysis reveals that this assumption is flawed: attention heads exhibit widely varying sensitivity to block granularity, and uniformity leads to suboptimal accuracy. We present AB-Sparse, a training-free algorithm-system co-designed framework that improves accuracy while preserving throughput. AB-Sparse introduces lightweight adaptive block size allocation across attention heads to improve accuracy. To compensate for the additional memory overhead, it further employs lossless block centroid quantization. In addition, custom GPU kernels are developed to support efficient execution with variable block sizes. Evaluation results demonstrate that AB-Sparse achieves an accuracy improvement of up to 5.43% over existing block sparse attention baselines without throughput overhead.

Large language models (LLMs) have demonstrated remarkable capabilities across various tasks. However, their widespread application is hindered by the resource-intensive decoding process. To address this challenge, current approaches have incorporated additional decoding heads to enable parallel prediction of multiple subsequent tokens, thereby achieving inference acceleration. Nevertheless, the accuracy of these decoding heads falls short of the auto-regressive decoding approach. In light of these limitations, we propose Chimera, a novel framework specifically designed for speculative sampling. Within this framework, we introduce a lightweight draft model that effectively utilizes previously generated tokens to predict subsequent words. To ensure both accuracy and efficiency, we present two strategies within the lightweight draft model. Firstly, we focus on capturing short-range dependencies at the bottom layer. Secondly, we leverage the readily available representations from the original LLM.Through empirical evaluation on the Vicuna and LlaMA-2 series, Chimera demonstrates impressive results, achieving an average latency speedup ratio of 2.7x compared to the vanilla auto-regressive decoding approach. This highlights the potential of our proposed framework in significantly improving the efficiency of large language models during the decoding process.

Diffusion models are widely recognized for generating high-quality and diverse images, but their poor real-time performance has led to numerous acceleration works, primarily focusing on UNet-based structures. With the more successful results achieved by diffusion transformers (DiT), there is still a lack of exploration regarding the impact of DiT structure on generation, as well as the absence of an acceleration framework tailored to the DiT architecture. To tackle these challenges, we conduct an investigation into the correlation between DiT blocks and image generation. Our findings reveal that the front blocks of DiT are associated with the outline of the generated images, while the rear blocks are linked to the details. Based on this insight, we propose an overall training-free inference acceleration framework Delta-DiT: using a designed cache mechanism to accelerate the rear DiT blocks in the early sampling stages and the front DiT blocks in the later stages. Specifically, a DiT-specific cache mechanism called Delta-Cache is proposed, which considers the inputs of the previous sampling image and reduces the bias in the inference. Extensive experiments on PIXART-alpha and DiT-XL demonstrate that the Delta-DiT can achieve a 1.6times speedup on the 20-step generation and even improves performance in most cases. In the scenario of 4-step consistent model generation and the more challenging 1.12times acceleration, our method significantly outperforms existing methods. Our code will be publicly available.

Diffusion Transformers (DiTs) have been widely used in many tasks, including image synthesis, video generation, and content editing. However, their multi-iteration inference process leads to performance inefficiency and high energy consumption. Existing acceleration methods primarily focus on reducing temporal redundancy between adjacent timesteps, but often overlook the specific features of DiTs. As a result, these approaches either suffer from great accuracy degradation or fail to achieve high efficiency. We present DSTAR, a software-hardware co-design framework that accelerates DiT inference by reducing spatial and temporal redundancy. At the algorithmic level, DSTAR introduces a fine-grained mixed-precision quantization method for differential activations in linear operations, significantly increasing the proportion of low-bit computations. Additionally, DSTAR incorporates a sparse attention reuse mechanism to minimize redundant computation in attention layers. For architectural support, we design a specialized hardware accelerator which achieves high efficiency in both latency and energy consumption. Evaluation on seven typical DiTs demonstrates that DSTAR achieves up to 7.33x latency speedup and 41.89x energy savings compared to an NVIDIA A100 GPU, and achieves up to 2.54x latency speedup and 3.68x energy savings compared to SOTA accelerators, without accuracy degradation.

Inference-time techniques are emerging as highly effective tools to enhance large language model (LLM) capabilities. However, best practices for developing systems that combine these techniques remain underdeveloped due to our limited understanding of the utility of individual inference-time techniques and the interactions between them. Additionally, efficiently and automatically searching the space of model choices, inference-time techniques, and their compositions is challenging due to the large design space. To address these challenges, we introduce Archon, a modular framework for selecting, combining, and stacking layers of inference-time techniques to construct optimized LLM systems for target benchmarks. Rather than relying on a single LLM called once, we leverage a diverse set of LLMs and inference-time techniques, creating LLM systems greater than the sum of their parts. Archon defines an extensible design space, encompassing techniques such as generation ensembling, repeated sampling, ranking, fusion, critiquing, verification, and unit testing. It transforms the problem of building LLM systems into a hyperparameter optimization objective. Given the available LLMs, inference-time techniques, and compute budget, Archon utilizes hyperparameter search techniques to discover optimized architectures for target benchmark(s). We evaluate Archon architectures across a range of instruction-following, reasoning, and coding benchmarks, including MT-Bench, Arena-Hard-Auto, AlpacaEval 2.0, MixEval, MixEval Hard, MATH, and CodeContests. Archon architectures outperform frontier models, such as GPT-4o and Claude 3.5 Sonnet, on these benchmarks, achieving an average accuracy increase of 15.1 percentage points by using all available LLMs. We make our code and datasets available publicly on Github: https://github.com/ScalingIntelligence/Archon.

Diffusion models produce impressive results in modalities ranging from images and video to protein design and text. However, generating samples with user-specified properties remains a challenge. Recent research proposes fine-tuning models to maximize rewards that capture desired properties, but these methods require expensive training and are prone to mode collapse. In this work, we propose Feynman Kac (FK) steering, an inference-time framework for steering diffusion models with reward functions. FK steering works by sampling a system of multiple interacting diffusion processes, called particles, and resampling particles at intermediate steps based on scores computed using functions called potentials. Potentials are defined using rewards for intermediate states and are selected such that a high value indicates that the particle will yield a high-reward sample. We explore various choices of potentials, intermediate rewards, and samplers. We evaluate FK steering on text-to-image and text diffusion models. For steering text-to-image models with a human preference reward, we find that FK steering a 0.8B parameter model outperforms a 2.6B parameter fine-tuned model on prompt fidelity, with faster sampling and no training. For steering text diffusion models with rewards for text quality and specific text attributes, we find that FK steering generates lower perplexity, more linguistically acceptable outputs and enables gradient-free control of attributes like toxicity. Our results demonstrate that inference-time scaling and steering of diffusion models, even with off-the-shelf rewards, can provide significant sample quality gains and controllability benefits. Code is available at https://github.com/zacharyhorvitz/Fk-Diffusion-Steering .

Source separation is a fundamental task in speech, music, and audio processing, and it also provides cleaner and larger data for training generative models. However, improving separation performance in practice often depends on increasingly large networks, inflating training and deployment costs. Motivated by recent advances in inference-time scaling for generative modeling, we propose Training-Time and Inference-Time Scalable Discriminative Source Separation (TISDiSS), a unified framework that integrates early-split multi-loss supervision, shared-parameter design, and dynamic inference repetitions. TISDiSS enables flexible speed-performance trade-offs by adjusting inference depth without retraining additional models. We further provide systematic analyses of architectural and training choices and show that training with more inference repetitions improves shallow-inference performance, benefiting low-latency applications. Experiments on standard speech separation benchmarks demonstrate state-of-the-art performance with a reduced parameter count, establishing TISDiSS as a scalable and practical framework for adaptive source separation. Code is available at https://github.com/WingSingFung/TISDiSS.

Large Language Models (LLMs) have sparked significant interest in their generative capabilities, leading to the development of various commercial applications. The high cost of using the models drives application builders to maximize the value of generation under a limited inference budget. This paper presents a study of optimizing inference hyperparameters such as the number of responses, temperature and max tokens, which significantly affects the utility/cost of text generation. We design a framework named EcoOptiGen which leverages economical hyperparameter optimization and cost-based pruning. Experiments with the GPT-3.5/GPT-4 models on a variety of tasks verify its effectiveness. EcoOptiGen is implemented in the `autogen' package of the FLAML library: https://aka.ms/autogen.

To fully leverage the capabilities of diffusion models, we are often interested in optimizing downstream reward functions during inference. While numerous algorithms for reward-guided generation have been recently proposed due to their significance, current approaches predominantly focus on single-shot generation, transitioning from fully noised to denoised states. We propose a novel framework for inference-time reward optimization with diffusion models inspired by evolutionary algorithms. Our approach employs an iterative refinement process consisting of two steps in each iteration: noising and reward-guided denoising. This sequential refinement allows for the gradual correction of errors introduced during reward optimization. Besides, we provide a theoretical guarantee for our framework. Finally, we demonstrate its superior empirical performance in protein and cell-type-specific regulatory DNA design. The code is available at https://github.com/masa-ue/ProDifEvo-Refinement{https://github.com/masa-ue/ProDifEvo-Refinement}.

Enabling private inference is crucial for many cloud inference services that are based on Transformer models. However, existing private inference solutions can increase the inference latency by more than 60x or significantly compromise the inference quality. In this paper, we design the framework MPCFORMER as a practical solution, using Secure Multi-Party Computation (MPC) and Knowledge Distillation (KD). Through extensive evaluations, we show that MPCFORMER significantly speeds up Transformer inference in MPC settings while achieving similar ML performance to the input model. On the IMDb dataset, it achieves similar performance to BERTBASE, while being 5.3x faster. On the GLUE benchmark, it achieves 97% performance of BERTBASE with a 2.2x speedup. MPCFORMER remains effective with different trained Transformer weights such as ROBERTABASE and larger models including BERTLarge. Code is available at https://github.com/MccRee177/MPCFormer.

As large language models become increasingly prevalent, memory bandwidth constraints significantly limit inference throughput, motivating post-training quantization (PTQ). In this paper, we propose FireQ, a co-designed PTQ framework and an INT4-FP8 matrix multiplication kernel that accelerates LLM inference across all linear layers. Specifically, FireQ quantizes linear layer weights and key-values to INT4, and activations and queries to FP8, significantly enhancing throughput. Additionally, we introduce a three-stage pipelining for the prefill phase, which modifies the FlashAttention-3 kernel, effectively reducing time-to-first-token in the prefill phase. To minimize accuracy loss from quantization, we develop novel outlier smoothing techniques tailored separately for linear and attention layers. In linear layers, we explicitly use per-tensor scaling to prevent underflow caused by the FP8 quantization scaling factor of INT4 quantization, and channel-wise scaling to compensate for coarse granularity of INT4. In attention layers, we address quantization challenges posed by rotary positional embeddings (RoPE) by combining pre-RoPE and post-RoPE scaling strategies. FireQ significantly outperforms state-of-the-art methods, achieving 1.68x faster inference in feed-forward network layers on Llama2-7B and 1.26x faster prefill phase performance on Llama3-8B compared to QServe, with negligible accuracy loss.

Machine learning on tiny IoT devices based on microcontroller units (MCU) is appealing but challenging: the memory of microcontrollers is 2-3 orders of magnitude smaller even than mobile phones. We propose MCUNet, a framework that jointly designs the efficient neural architecture (TinyNAS) and the lightweight inference engine (TinyEngine), enabling ImageNet-scale inference on microcontrollers. TinyNAS adopts a two-stage neural architecture search approach that first optimizes the search space to fit the resource constraints, then specializes the network architecture in the optimized search space. TinyNAS can automatically handle diverse constraints (i.e.device, latency, energy, memory) under low search costs.TinyNAS is co-designed with TinyEngine, a memory-efficient inference library to expand the search space and fit a larger model. TinyEngine adapts the memory scheduling according to the overall network topology rather than layer-wise optimization, reducing the memory usage by 4.8x, and accelerating the inference by 1.7-3.3x compared to TF-Lite Micro and CMSIS-NN. MCUNet is the first to achieves >70% ImageNet top1 accuracy on an off-the-shelf commercial microcontroller, using 3.5x less SRAM and 5.7x less Flash compared to quantized MobileNetV2 and ResNet-18. On visual&audio wake words tasks, MCUNet achieves state-of-the-art accuracy and runs 2.4-3.4x faster than MobileNetV2 and ProxylessNAS-based solutions with 3.7-4.1x smaller peak SRAM. Our study suggests that the era of always-on tiny machine learning on IoT devices has arrived. Code and models can be found here: https://tinyml.mit.edu.

Novel View Synthesis (NVS) is the task of generating new images of a scene from viewpoints that were not part of the original input. Diffusion-based NVS can generate high-quality, temporally consistent images, however, remains computationally prohibitive. Conversely, regression-based NVS offers suboptimal generation quality despite requiring significantly lower compute; leaving the design objective of a high-quality, inference-efficient NVS framework an open challenge. To close this critical gap, we present Sphinx, a training-free hybrid inference framework that achieves diffusion-level fidelity at a significantly lower compute. Sphinx proposes to use regression-based fast initialization to guide and reduce the denoising workload for the diffusion model. Additionally, it integrates selective refinement with adaptive noise scheduling, allowing more compute to uncertain regions and frames. This enables Sphinx to provide flexible navigation of the performance-quality trade-off, allowing adaptation to latency and fidelity requirements for dynamically changing inference scenarios. Our evaluation shows that Sphinx achieves an average 1.8x speedup over diffusion model inference with negligible perceptual degradation of less than 5%, establishing a new Pareto frontier between quality and latency in NVS serving.

Autoregressive models, despite their commendable performance in a myriad of generative tasks, face challenges stemming from their inherently sequential structure. Inference on these models, by design, harnesses a temporal dependency, where the current token's probability distribution is conditioned on preceding tokens. This inherent characteristic severely impedes computational efficiency during inference as a typical inference request can require more than thousands of tokens, where generating each token requires a load of entire model weights, making the inference more memory-bound. The large overhead becomes profound in real deployment where requests arrive randomly, necessitating various generation lengths. Existing solutions, such as dynamic batching and concurrent instances, introduce significant response delays and bandwidth contention, falling short of achieving optimal latency and throughput. To address these shortcomings, we propose Flover -- a temporal fusion framework for efficiently inferring multiple requests in parallel. We deconstruct the general generation pipeline into pre-processing and token generation, and equip the framework with a dedicated work scheduler for fusing the generation process temporally across all requests. By orchestrating the token-level parallelism, Flover exhibits optimal hardware efficiency and significantly spares the system resources. By further employing a fast buffer reordering algorithm that allows memory eviction of finished tasks, it brings over 11x inference speedup on GPT and 16x on LLAMA compared to the cutting-edge solutions provided by NVIDIA FasterTransformer. Crucially, by leveraging the advanced tensor parallel technique, Flover proves efficacious across diverse computational landscapes, from single-GPU setups to distributed scenarios, thereby offering robust performance optimization that adapts to variable use cases.

Striking an optimal balance between minimal drafting latency and high speculation accuracy to enhance the inference speed of Large Language Models remains a significant challenge in speculative decoding. In this paper, we introduce Falcon, an innovative semi-autoregressive speculative decoding framework fashioned to augment both the drafter's parallelism and output quality. Falcon incorporates the Coupled Sequential Glancing Distillation technique, which fortifies inter-token dependencies within the same block, leading to increased speculation accuracy. We offer a comprehensive theoretical analysis to illuminate the underlying mechanisms. Additionally, we introduce a Custom-Designed Decoding Tree, which permits the drafter to generate multiple tokens in a single forward pass and accommodates multiple forward passes as needed, thereby boosting the number of drafted tokens and significantly improving the overall acceptance rate. Comprehensive evaluations on benchmark datasets such as MT-Bench, HumanEval, and GSM8K demonstrate Falcon's superior acceleration capabilities. The framework achieves a lossless speedup ratio ranging from 2.91x to 3.51x when tested on the Vicuna and LLaMA2-Chat model series. These results outstrip existing speculative decoding methods for LLMs, including Eagle, Medusa, Lookahead, SPS, and PLD, while maintaining a compact drafter architecture equivalent to merely two Transformer layers.

Achieving precise alignment between user intent and generated visuals remains a central challenge in text-to-visual generation, as a single attempt often fails to produce the desired output. To handle this, prior approaches mainly scale the visual generation process (e.g., increasing sampling steps or seeds), but this quickly leads to a quality plateau. This limitation arises because the prompt, crucial for guiding generation, is kept fixed. To address this, we propose Prompt Redesign for Inference-time Scaling, coined PRIS, a framework that adaptively revises the prompt during inference in response to the scaled visual generations. The core idea of PRIS is to review the generated visuals, identify recurring failure patterns across visuals, and redesign the prompt accordingly before regenerating the visuals with the revised prompt. To provide precise alignment feedback for prompt revision, we introduce a new verifier, element-level factual correction, which evaluates the alignment between prompt attributes and generated visuals at a fine-grained level, achieving more accurate and interpretable assessments than holistic measures. Extensive experiments on both text-to-image and text-to-video benchmarks demonstrate the effectiveness of our approach, including a 15% gain on VBench 2.0. These results highlight that jointly scaling prompts and visuals is key to fully leveraging scaling laws at inference-time. Visualizations are available at the website: https://subin-kim-cv.github.io/PRIS.

Transformer-based large models excel in natural language processing and computer vision, but face severe computational inefficiencies due to the self-attention's quadratic complexity with input tokens. Recently, researchers have proposed a series of methods based on block selection and compression to alleviate this problem, but they either have issues with semantic incompleteness or poor training-inference efficiency. To comprehensively address these challenges, we propose ChunkLLM, a lightweight and pluggable training framework. Specifically, we introduce two components: QK Adapter (Q-Adapter and K-Adapter) and Chunk Adapter. The former is attached to each Transformer layer, serving dual purposes of feature compression and chunk attention acquisition. The latter operates at the bottommost layer of the model, functioning to detect chunk boundaries by leveraging contextual semantic information. During the training phase, the parameters of the backbone remain frozen, with only the QK Adapter and Chunk Adapter undergoing training. Notably, we design an attention distillation method for training the QK Adapter, which enhances the recall rate of key chunks. During the inference phase, chunk selection is triggered exclusively when the current token is detected as a chunk boundary, thereby accelerating model inference. Experimental evaluations are conducted on a diverse set of long-text and short-text benchmark datasets spanning multiple tasks. ChunkLLM not only attains comparable performance on short-text benchmarks but also maintains 98.64% of the performance on long-context benchmarks while preserving a 48.58% key-value cache retention rate. Particularly, ChunkLLM attains a maximum speedup of 4.48x in comparison to the vanilla Transformer in the processing of 120K long texts.

Deploying Large Language Models (LLMs) on edge devices remains challenging due to their quadratically increasing computations with the sequence length. Existing studies for dynamic attention pruning are designed for hardware with massively parallel computation capabilities, such as GPUs or TPUs, and aim at long context lengths (e.g., 64K), making them unsuitable for edge scenarios. We present DeltaLLM, a training-free framework that exploits temporal sparsity in attention patterns to enable efficient LLM inference across both the prefilling and decoding stages, on resource-constrained edge devices. DeltaLLM introduces an accuracy- and memory-aware delta matrix construction strategy that introduces temporal sparsity, and a context-aware hybrid attention mechanism that combines full attention in a local context window with delta approximation outside it to increase accuracy. We evaluate our framework on the edge-device-friendly BitNet-b1.58-2B-4T model and Llama3.2-1B-Instruct model across diverse language tasks. The results show that on BitNet, our framework increases the attention sparsity from 0% to 60% during the prefilling stage with slight accuracy improvement on the WG task, and 0% to 57% across both the prefilling and decoding stages, with even higher F1 score from 29.63 to 30.97 on SQuAD-v2 task. On the Llama model, it can also achieve up to 60% sparsity during the prefilling stage and around 57% across both stages with negligible accuracy drop. These results demonstrate that DeltaLLM offers a promising solution for efficient edge deployment, requiring no fine-tuning and seamlessly integrating with existing inference pipelines.

Point cloud analysis is challenging due to irregularity and unordered data structure. To capture the 3D geometries, prior works mainly rely on exploring sophisticated local geometric extractors using convolution, graph, or attention mechanisms. These methods, however, incur unfavorable latency during inference, and the performance saturates over the past few years. In this paper, we present a novel perspective on this task. We notice that detailed local geometrical information probably is not the key to point cloud analysis -- we introduce a pure residual MLP network, called PointMLP, which integrates no sophisticated local geometrical extractors but still performs very competitively. Equipped with a proposed lightweight geometric affine module, PointMLP delivers the new state-of-the-art on multiple datasets. On the real-world ScanObjectNN dataset, our method even surpasses the prior best method by 3.3% accuracy. We emphasize that PointMLP achieves this strong performance without any sophisticated operations, hence leading to a superior inference speed. Compared to most recent CurveNet, PointMLP trains 2x faster, tests 7x faster, and is more accurate on ModelNet40 benchmark. We hope our PointMLP may help the community towards a better understanding of point cloud analysis. The code is available at https://github.com/ma-xu/pointMLP-pytorch.

Graphic layouts serve as an important and engaging medium for visual communication across different channels. While recent layout generation models have demonstrated impressive capabilities, they frequently fail to align with nuanced human aesthetic judgment. Existing preference datasets and reward models trained on text-to-image generation do not generalize to layout evaluation, where the spatial arrangement of identical elements determines quality. To address this critical gap, we introduce DesignSense-10k, a large-scale dataset of 10,235 human-annotated preference pairs for graphic layout evaluation. We propose a five-stage curation pipeline that generates visually coherent layout transformations across diverse aspect ratios, using semantic grouping, layout prediction, filtering, clustering, and VLM-based refinement to produce high-quality comparison pairs. Human preferences are annotated using a 4-class scheme (left, right, both good, both bad) to capture subjective ambiguity. Leveraging this dataset, we train DesignSense, a vision-language model-based classifier that substantially outperforms existing open-source and proprietary models across comprehensive evaluation metrics (54.6% improvement in Macro F1 over the strongest proprietary baseline). Our analysis shows that frontier VLMs remain unreliable overall and fail catastrophically on the full four-class task, underscoring the need for specialized, preference-aware models. Beyond the dataset, our reward model DesignSense yields tangible downstream gains in layout generation. Using our judge during RL based training improves generator win rate by about 3%, while inference-time scaling, which involves generating multiple candidates and selecting the best one, provides a 3.6% improvement. These results highlight the practical impact of specialized, layout-aware preference modeling on real-world layout generation quality.

Serving large language models (LLMs) in production can incur substantial costs, which has prompted recent advances in inference system optimizations. Today, these systems are evaluated against conventional latency and throughput metrics (eg. TTFT, TBT, Normalised Latency and TPOT). However, these metrics fail to fully capture the nuances of LLM inference, leading to an incomplete assessment of user-facing performance crucial for real-time applications such as chat and translation. In this paper, we first identify the pitfalls of current performance metrics in evaluating LLM inference systems. We then propose Etalon, a comprehensive performance evaluation framework that includes fluidity-index -- a novel metric designed to reflect the intricacies of the LLM inference process and its impact on real-time user experience. Finally, we evaluate various existing open-source platforms and model-as-a-service offerings using Etalon, discussing their strengths and weaknesses. Etalon is available at https://github.com/project-etalon/etalon.

Hybrid model architectures that combine computational primitives (e.g., Attention, MLP) in different ratios have shown promising performance beyond Transformers. Some studies have shown that different interleavings of primitives can affect model quality as well. However, prior works explore the hybrid model architecture design space manually. Due to the large design space and training costs, discovering hybrid models that combine key computational primitives for pre-training is challenging. In this work, we take a principled approach in designing a modular hybrid model architecture search framework -- Composer. Composer explores model architectures at a small scale and extrapolates the top-performing model architectures to a larger scale using our proposed scaling strategies. Using Composer, we discover new hybrid LLM architectures that outperform Llama 3.2. Compared to Llama 3.2 and previous state-of-the-art baselines, the new model architectures consistently reduce validation loss at parameter scales of 350M-3B and improve evaluation accuracy on the downstream tasks by up to 2.8-8.3% (1.1-3.1% on average) while improving both training and inference efficiency.

The emergence of LLMs has catalyzed a paradigm shift in autonomous agent development, enabling systems capable of reasoning, planning, and executing complex multi-step tasks. However, existing agent frameworks often suffer from architectural rigidity, vendor lock-in, and prohibitive complexity that impedes rapid prototyping and deployment. This paper presents AgentForge, a lightweight, open-source Python framework designed to democratize the construction of LLM-driven autonomous agents through a principled modular architecture. AgentForge introduces three key innovations: (1) a composable skill abstraction that enables fine-grained task decomposition with formally defined input-output contracts, (2) a unified LLM backend interface supporting seamless switching between cloud-based APIs and local inference engines, and (3) a declarative YAML-based configuration system that separates agent logic from implementation details. We formalize the skill composition mechanism as a directed acyclic graph (DAG) and prove its expressiveness for representing arbitrary sequential and parallel task workflows. Comprehensive experimental evaluation across four benchmark scenarios demonstrates that AgentForge achieves competitive task completion rates while reducing development time by 62% compared to LangChain and 78% compared to direct API integration. Latency measurements confirm sub-100ms orchestration overhead, rendering the framework suitable for real-time applications. The modular design facilitates extension: we demonstrate the integration of six built-in skills and provide comprehensive documentation for custom skill development. AgentForge addresses a critical gap in the LLM agent ecosystem by providing researchers and practitioners with a production-ready foundation for constructing, evaluating, and deploying autonomous agents without sacrificing flexibility or performance.

Recently, large language models (LLMs) have achieved huge success in the natural language processing (NLP) field, driving a growing demand to extend their deployment from the cloud to edge devices. However, deploying LLMs on resource-constrained edge devices poses significant challenges, including (1) intensive computations and huge model sizes, (2) great memory and bandwidth demands introduced by the autoregressive generation process, and (3) limited scalability for handling long sequences. To address these challenges, we propose AccLLM, a comprehensive acceleration framework that enables efficient and fast long-context LLM inference through algorithm and hardware co-design. At the algorithmic level, we integrate (1) pruning, (2) {\Lambda}-shaped attention, and (3) an innovative W2A8KV4 (2-bit weights, 8-bit activations, and 4-bit KV cache) quantization scheme, thus effectively reducing memory and bandwidth requirements while facilitating LLMs' long-sequence generation. At the hardware level, we design a dedicated FPGA-based accelerator with a reconfigurable computing engine to effectively and flexibly accommodate diverse operations arising from our compression algorithm, thereby fully translating the algorithmic innovations into tangible hardware efficiency. We validate AccLLM on the Xilinx Alveo U280 FPGA, demonstrating a 4.07x energy efficiency and a 2.98x throughput compared to the state-of-the-art work FlightLLM.

The integration of extensive, dynamic knowledge into Large Language Models (LLMs) remains a significant challenge due to the inherent entanglement of factual data and reasoning patterns. Existing solutions, ranging from non-parametric Retrieval-Augmented Generation (RAG) to parametric knowledge editing, are often constrained in practice by finite context windows, retriever noise, or the risk of catastrophic forgetting. In this paper, we propose DRIFT, a novel dual-model architecture designed to explicitly decouple knowledge extraction from the reasoning process. Unlike static prompt compression, DRIFT employs a lightweight knowledge model to dynamically compress document chunks into implicit fact tokens conditioned on the query. These dense representations are projected into the reasoning model's embedding space, replacing raw, redundant text while maintaining inference accuracy. Extensive experiments show that DRIFT significantly improves performance on long-context tasks, outperforming strong baselines among comparably sized models. Our approach provides a scalable and efficient paradigm for extending the effective context window and reasoning capabilities of LLMs. Our code is available at https://github.com/Lancelot-Xie/DRIFT.

This paper presents EasyRAG, a simple, lightweight, and efficient retrieval-augmented generation framework for automated network operations. Our framework has three advantages. The first is accurate question answering. We designed a straightforward RAG scheme based on (1) a specific data processing workflow (2) dual-route sparse retrieval for coarse ranking (3) LLM Reranker for reranking (4) LLM answer generation and optimization. This approach achieved first place in the GLM4 track in the preliminary round and second place in the GLM4 track in the semifinals. The second is simple deployment. Our method primarily consists of BM25 retrieval and BGE-reranker reranking, requiring no fine-tuning of any models, occupying minimal VRAM, easy to deploy, and highly scalable; we provide a flexible code library with various search and generation strategies, facilitating custom process implementation. The last one is efficient inference. We designed an efficient inference acceleration scheme for the entire coarse ranking, reranking, and generation process that significantly reduces the inference latency of RAG while maintaining a good level of accuracy; each acceleration scheme can be plug-and-play into any component of the RAG process, consistently enhancing the efficiency of the RAG system. Our code and data are released at https://github.com/BUAADreamer/EasyRAG.

We present a novel framework for Industry 5.0 that simplifies the deployment of AI models on edge devices in various industrial settings. The design reduces latency and avoids external data transfer by enabling local inference and real-time processing. Our implementation is agent-based, which means that individual agents, whether human, algorithmic, or collaborative, are responsible for well-defined tasks, enabling flexibility and simplifying integration. Moreover, our framework supports modular integration and maintains low resource requirements. Preliminary evaluations concerning the food industry in real scenarios indicate improved deployment time and system adaptability performance. The source code is publicly available at https://github.com/AI-REDGIO-5-0/ci-component.

Security of model parameters and user data is critical for Transformer-based services, such as ChatGPT. While recent strides in secure two-party protocols have successfully addressed security concerns in serving Transformer models, their adoption is practically infeasible due to the prohibitive cryptographic overheads involved. Drawing insights from our hands-on experience in developing two real-world Transformer-based services, we identify the inherent efficiency bottleneck in the two-party assumption. To overcome this limitation, we propose a novel three-party threat model. Within this framework, we design a semi-symmetric permutation-based protection scheme and present STIP, the first secure Transformer inference protocol without any inference accuracy loss. Experiments on representative Transformer models in real systems show that STIP has practical security and outperforms state-of-the-art secure two-party protocols in efficiency by millions of times.

Due to increasing privacy concerns, neural network (NN) based secure inference (SI) schemes that simultaneously hide the client inputs and server models attract major research interests. While existing works focused on developing secure protocols for NN-based SI, in this work, we take a different approach. We propose NASS, an integrated framework to search for tailored NN architectures designed specifically for SI. In particular, we propose to model cryptographic protocols as design elements with associated reward functions. The characterized models are then adopted in a joint optimization with predicted hyperparameters in identifying the best NN architectures that balance prediction accuracy and execution efficiency. In the experiment, it is demonstrated that we can achieve the best of both worlds by using NASS, where the prediction accuracy can be improved from 81.6% to 84.6%, while the inference runtime is reduced by 2x and communication bandwidth by 1.9x on the CIFAR-10 dataset.

Discrete denoising diffusion models (DDMs) have recently emerged as a compelling alternative to autoregressive (AR) modeling for discrete data, offering parallel generation and iterative global refinement capabilities. Unlike continuous diffusion, where the state space is fixed, DDMs are fundamentally shaped by how the discrete state space is constructed: the tokenization scheme, the vocabulary topology, and domain-specific structural alphabets. This work introduces a unified conceptual framework that views discrete diffusion models through the construction of the underlying discrete state space. Within this framework, existing formulations, including transition-matrix, masking/absorbing-state, and score/ratio-based approaches, emerge as different instantiations of a common design space. The framework further exposes common design trade-offs across training objectives, inference algorithms, scaling behavior, systems optimization, and evaluation protocols, suggesting several promising directions for future research.

We revisit retrieval-augmented generation (RAG) by embedding retrieval control directly into generation. Instead of treating retrieval as an external intervention, we express retrieval decisions within token-level decoding, enabling end-to-end coordination without additional controllers or classifiers. Under the paradigm of Retrieval as Generation, we propose GRIP (Generation-guided Retrieval with Information Planning), a unified framework in which the model regulates retrieval behavior through control-token emission. Central to GRIP is Self-Triggered Information Planning, which allows the model to decide when to retrieve, how to reformulate queries, and when to terminate, all within a single autoregressive trajectory. This design tightly couples retrieval and reasoning and supports dynamic multi-step inference with on-the-fly evidence integration. To supervise these behaviors, we construct a structured training set covering answerable, partially answerable, and multi-hop queries, each aligned with specific token patterns. Experiments on five QA benchmarks show that GRIP surpasses strong RAG baselines and is competitive with GPT-4o while using substantially fewer parameters.

With the rapid development of large language models (LLMs), distributed training and inference frameworks like DeepSpeed have become essential for scaling model training and inference across multiple GPUs or nodes. However, the increasing complexity of these frameworks brings non-trivial software bugs, which may degrade training performance, cause unexpected failures, and result in significant resource waste. Understanding framework bugs' characteristics is fundamental for quality assurance, allowing the design of more effective debugging and repair methods. Thus, our paper conducts the first large-scale empirical analysis of 308 fixed bugs across three popular distributed training/inference frameworks: DeepSpeed, Megatron-LM, and Colossal-AI. We examine bug symptoms, root causes, bug identification and fixing efforts, and common low-effort fixing strategies. Additionally, the distributed nature of these frameworks introduces unique bug root causes, such as allocation strategy error and distributed communication error. Diagnosing and fixing complex bugs remains challenging due to factors like the disconnect between symptoms and root causes, high bug reproduction costs, and low-level or cross-component interactions. Interestingly, we observe that 48% of bug fixes require minimal code changes (<=10 LOC) and follow simple strategies such as conditional logic optimization, parameter handling enhancement, or version compatibility handling, indicating potential for automation. Based on these insights, we offer several implications for improving the reliability of both distributed training and inference frameworks and their dependent LLM projects, while also identifying opportunities to leverage LLM-based tools for automated debugging and repair.

Audio-driven human animation technology is widely used in human-computer interaction, and the emergence of diffusion models has further advanced its development. Currently, most methods rely on multi-stage generation and intermediate representations, resulting in long inference time and issues with generation quality in specific foreground regions and audio-motion consistency. These shortcomings are primarily due to the lack of localized fine-grained supervised guidance. To address above challenges, we propose Parts-aware Audio-driven Human Animation, PAHA, a unit enhancement and guidance framework for audio-driven upper-body animation. We introduce two key methods: Parts-Aware Re-weighting (PAR) and Parts Consistency Enhancement (PCE). PAR dynamically adjusts regional training loss weights based on pose confidence scores, effectively improving visual quality. PCE constructs and trains diffusion-based regional audio-visual classifiers to improve the consistency of motion and co-speech audio. Afterwards, we design two novel inference guidance methods for the foregoing classifiers, Sequential Guidance (SG) and Differential Guidance (DG), to balance efficiency and quality respectively. Additionally, we build CNAS, the first public Chinese News Anchor Speech dataset, to advance research and validation in this field. Extensive experimental results and user studies demonstrate that PAHA significantly outperforms existing methods in audio-motion alignment and video-related evaluations. The codes and CNAS dataset will be released upon acceptance.

We present a comprehensive framework for enhancing Retrieval-Augmented Generation (RAG) systems through dynamic retrieval strategies and reinforcement fine-tuning. This approach significantly improves large language models on knowledge-intensive tasks, including opendomain question answering and complex reasoning. Our framework integrates two complementary techniques: Policy-Optimized RetrievalAugmented Generation (PORAG), which optimizes the use of retrieved information, and Adaptive Token-Layer Attention Scoring (ATLAS), which dynamically determines retrieval timing and content based on contextual needs. Together, these techniques enhance both the utilization and relevance of retrieved content, improving factual accuracy and response quality. Designed as a lightweight solution compatible with any Transformer-based LLM without requiring additional training, our framework excels in knowledge-intensive tasks, boosting output accuracy in RAG settings. We further propose CRITIC, a novel method to selectively compress key-value caches by token importance, mitigating memory bottlenecks in long-context applications. The framework also incorporates test-time scaling techniques to dynamically balance reasoning depth and computational resources, alongside optimized decoding strategies for faster inference. Experiments on benchmark datasets show that our framework reduces hallucinations, strengthens domain-specific reasoning, and achieves significant efficiency and scalability gains over traditional RAG systems. This integrated approach advances the development of robust, efficient, and scalable RAG systems across diverse applications.

Effective interaction modeling and behavior prediction of dynamic agents play a significant role in interactive motion planning for autonomous robots. Although existing methods have improved prediction accuracy, few research efforts have been devoted to enhancing prediction model interpretability and out-of-distribution (OOD) generalizability. This work addresses these two challenging aspects by designing a variational auto-encoder framework that integrates graph-based representations and time-sequence models to efficiently capture spatio-temporal relations between interactive agents and predict their dynamics. Our model infers dynamic interaction graphs in a latent space augmented with interpretable edge features that characterize the interactions. Moreover, we aim to enhance model interpretability and performance in OOD scenarios by disentangling the latent space of edge features, thereby strengthening model versatility and robustness. We validate our approach through extensive experiments on both simulated and real-world datasets. The results show superior performance compared to existing methods in modeling spatio-temporal relations, motion prediction, and identifying time-invariant latent features.

Urban causal research is essential for understanding the complex dynamics of cities and informing evidence-based policies. However, it is challenged by the inefficiency and bias of hypothesis generation, barriers to multimodal data complexity, and the methodological fragility of causal experimentation. Recent advances in large language models (LLMs) present an opportunity to rethink how urban causal analysis is conducted. This Perspective examines current urban causal research by analyzing taxonomies that categorize research topics, data sources, and methodological approaches to identify structural gaps. We then introduce an LLM-driven conceptual framework, AutoUrbanCI, composed of four distinct modular agents responsible for hypothesis generation, data engineering, experiment design and execution, and results interpretation with policy recommendations. We propose evaluation criteria for rigor and transparency and reflect on implications for human-AI collaboration, equity, and accountability. We call for a new research agenda that embraces AI-augmented workflows not as replacements for human expertise but as tools to broaden participation, improve reproducibility, and unlock more inclusive forms of urban causal reasoning.

Self-supervised monocular depth estimation has gathered notable interest since it can liberate training from dependency on depth annotations. In monocular video training case, recent methods only conduct view synthesis between existing camera views, leading to insufficient guidance. To tackle this, we try to synthesize more virtual camera views by flow-based video frame interpolation (VFI), termed as temporal augmentation. For multi-frame inference, to sidestep the problem of dynamic objects encountered by explicit geometry-based methods like ManyDepth, we return to the feature fusion paradigm and design a VFI-assisted multi-frame fusion module to align and aggregate multi-frame features, using motion and occlusion information obtained by the flow-based VFI model. Finally, we construct a unified self-supervised learning framework, named Mono-ViFI, to bilaterally connect single- and multi-frame depth. In this framework, spatial data augmentation through image affine transformation is incorporated for data diversity, along with a triplet depth consistency loss for regularization. The single- and multi-frame models can share weights, making our framework compact and memory-efficient. Extensive experiments demonstrate that our method can bring significant improvements to current advanced architectures. Source code is available at https://github.com/LiuJF1226/Mono-ViFI.

Ternary weight quantization (e.g., BitNet b1.58) offers a promising path to mitigate the memory bandwidth bottleneck in Large Language Model (LLM) inference. However, conventional compute platforms lack native support for ternary-weight arithmetic, often relying on inefficient dequantization. Lookup table (LUT)-based hardware architectures provide an effective alternative by replacing multiplications with conditional additions, but their design space remains largely unexplored. Existing designs rely on heuristic parameter selection, lacking a systematic understanding of the architectural trade-offs. This work addresses this gap by formalizing the design space of ternary LUT-based accelerators and presenting an open-source hardware generator coupled with an analytical cost model, validated against synthesis in TSMC 16nm technology. By spanning the full architectural space, this framework not only enables rapid design space exploration but also establishes a common footing for fair cross-design evaluation, which was previously hindered by inconsistent instantiations across published accelerators. Using this framework, we challenge several assumptions and design choices in recent literature. We demonstrate that the optimal architecture is fundamentally governed by the activation data type: while LUT-based reuse offers significant gains for high-cost arithmetic (e.g., FP16), it yields diminishing returns for small integer types. Furthermore, we show that maximizing core size consistently improves area density compared to highly tiled approaches. Our optimized designs achieve a 2.2x area reduction compared to multiplier-based baselines. Moreover, by benchmarking state-of-the-art implementations against our model, we reveal that correcting suboptimal parameters yields up to a 1.2x area improvement.

Large Language Models (LLMs) demonstrate exceptional performance but entail significant memory and computational costs, restricting their practical deployment. While existing INT4/INT8 quantization reduces these costs, they often degrade accuracy or lack optimal efficiency. INT6 quantization offers a superior trade-off between model accuracy and inference efficiency, but lacks hardware support in modern GPUs, forcing emulation via higher-precision arithmetic units that limit acceleration. In this paper, we propose FlexQ, a novel post-training INT6 quantization framework combining algorithmic innovation with system-level optimizations. FlexQ employs uniform 6-bit weight quantization across all layers, with adaptive retention of 8-bit activations in layers identified through layer-wise sensitivity analysis. To maximize hardware efficiency, we develop a specialized high-performance GPU kernel supporting matrix multiplication for W6A6 and W6A8 representations via Binary Tensor Core (BTC) equivalents, effectively bypassing the lack of native INT6 tensor cores. Evaluations on LLaMA models show FlexQ maintains near-FP16 accuracy, with perplexity increases of no more than 0.05. The proposed kernel achieves an average 1.39times speedup over ABQ-LLM on LLaMA-2-70B linear layers. End-to-end, FlexQ delivers 1.33times inference acceleration and 1.21times memory savings over SmoothQuant. Code is released at https://github.com/FlyFoxPlayer/FlexQ.

We introduce a gradient-free framework for Bayesian Optimal Experimental Design (BOED) in sequential settings, aimed at complex systems where gradient information is unavailable. Our method combines Ensemble Kalman Inversion (EKI) for design optimization with the Affine-Invariant Langevin Dynamics (ALDI) sampler for efficient posterior sampling-both of which are derivative-free and ensemble-based. To address the computational challenges posed by nested expectations in BOED, we propose variational Gaussian and parametrized Laplace approximations that provide tractable upper and lower bounds on the Expected Information Gain (EIG). These approximations enable scalable utility estimation in high-dimensional spaces and PDE-constrained inverse problems. We demonstrate the performance of our framework through numerical experiments ranging from linear Gaussian models to PDE-based inference tasks, highlighting the method's robustness, accuracy, and efficiency in information-driven experimental design.

The proliferation of large language models (LLMs) has led to the adoption of Mixture-of-Experts (MoE) architectures that dynamically leverage specialized subnetworks for improved efficiency and performance. Despite their benefits, MoE models face significant challenges during inference, including inefficient memory management and suboptimal batching, due to misaligned design choices between the model architecture and the system policies. Furthermore, the conventional approach of training MoEs from scratch is increasingly prohibitive in terms of cost. In this paper, we propose a novel framework Read-ME that transforms pre-trained dense LLMs into smaller MoE models (in contrast to "upcycling" generalist MoEs), avoiding the high costs of ground-up training. Our approach employs activation sparsity to extract experts. To compose experts, we examine the widely-adopted layer-wise router design and show its redundancy, and thus we introduce the pre-gating router decoupled from the MoE backbone that facilitates system-friendly pre-computing and lookahead scheduling, enhancing expert-aware batching and caching. Our codesign therefore addresses critical gaps on both the algorithmic and system fronts, establishing a scalable and efficient alternative for LLM inference in resource-constrained settings. Read-ME outperforms other popular open-source dense models of similar scales, achieving improvements of up to 10.1% on MMLU, and improving mean end-to-end latency up to 6.1%. Codes are available at: https://github.com/VITA-Group/READ-ME.

Efficient parallelization of Large Language Models (LLMs) with long sequences is essential but challenging due to their significant computational and memory demands, particularly stemming from communication bottlenecks in attention mechanisms. While sequence parallelism (SP) has been introduced as a potential solution, existing methods often suffer from limited scalability or inefficiency, rendering their effectiveness. Ring-Attention demonstrates the potential for scaling sequence processing but faces significant limitations due to its reliance on peer-to-peer (P2P) communication and inefficient utilization of network resources. As the degree of SP increases, the quadratic decrease in computation time per step contrasts sharply with the linear reduction in communication volume, exacerbating communication bottlenecks. To address these challenges, we propose TokenRing, a fine-grained parallel framework that leverages bidirectional P2P communication to effectively overlap computation and data transmission. By partitioning the attention block and concurrently transmitting Query and block outputs (i.e., block_out and block_lse) within a fully connected mesh topology, TokenRing achieves significant reductions in communication overhead and better load balancing. These innovations improve the scalability and efficiency of distributed Transformer models, particularly for long-context sequences. Experimental results demonstrate that TokenRing enhances throughput and reduces communication latency. Moreover, its design adapts seamlessly to various multi-GPU interconnect solutions, such as Huawei Ascend, ensuring broad compatibility and cost-effectiveness for distributed LLM inference and training. The code is available at: https://github.com/ACA-Lab-SJTU/token-ring.

Recent progress in vision and language foundation models has significantly advanced multimodal understanding, reasoning, and generation, inspiring a surge of interest in extending such capabilities to embodied settings through vision-language-action (VLA) models. Yet, most VLA models are still trained with supervised fine-tuning (SFT), which struggles to generalize under distribution shifts due to error accumulation. Reinforcement learning (RL) offers a promising alternative by directly optimizing task performance through interaction, but existing attempts remain fragmented and lack a unified platform for fair and systematic comparison across model architectures and algorithmic designs. To address this gap, we introduce RLinf-VLA, a unified and efficient framework for scalable RL training of VLA models. The system adopts a highly flexible resource allocation design that addresses the challenge of integrating rendering, training, and inference in RL+VLA training. In particular, for GPU-parallelized simulators, RLinf-VLA implements a novel hybrid fine-grained pipeline allocation mode, achieving a 1.61x-1.88x speedup in training. Through a unified interface, RLinf-VLA seamlessly supports diverse VLA architectures (e.g., OpenVLA, OpenVLA-OFT), multiple RL algorithms (e.g., PPO, GRPO), and various simulators (e.g., ManiSkill, LIBERO). In simulation, a unified model achieves 98.11\% across 130 LIBERO tasks and 97.66\% across 25 ManiSkill tasks. Beyond empirical performance, our study distills a set of best practices for applying RL to VLA training and sheds light on emerging patterns in this integration. Furthermore, we present preliminary deployment on a real-world Franka robot, where RL-trained policies exhibit stronger generalization than those trained with SFT. We envision RLinf-VLA as a foundation to accelerate and standardize research on embodied intelligence.

With the fast growth of parameter size, it becomes increasingly challenging to deploy large generative models as they typically require large GPU memory consumption and massive computation. Unstructured model pruning has been a common approach to reduce both GPU memory footprint and the overall computation while retaining good model accuracy. However, the existing solutions do not provide a highly-efficient support for handling unstructured sparsity on modern GPUs, especially on the highly-structured Tensor Core hardware. Therefore, we propose Flash-LLM for enabling low-cost and highly-efficient large generative model inference with the sophisticated support of unstructured sparsity on high-performance but highly restrictive Tensor Cores. Based on our key observation that the main bottleneck of generative model inference is the several skinny matrix multiplications for which Tensor Cores would be significantly under-utilized due to low computational intensity, we propose a general Load-as-Sparse and Compute-as-Dense methodology for unstructured sparse matrix multiplication. The basic insight is to address the significant memory bandwidth bottleneck while tolerating redundant computations that are not critical for end-to-end performance on Tensor Cores. Based on this, we design an effective software framework for Tensor Core based unstructured SpMM, leveraging on-chip resources for efficient sparse data extraction and computation/memory-access overlapping. At SpMM kernel level, Flash-LLM significantly outperforms the state-of-the-art library, i.e., Sputnik and SparTA by an average of 2.9x and 1.5x, respectively. At end-to-end framework level on OPT-30B/66B/175B models, for tokens per GPU-second, Flash-LLM achieves up to 3.8x and 3.6x improvement over DeepSpeed and FasterTransformer, respectively, with significantly lower inference cost.

Conditional generative models typically demand large annotated training sets to achieve high-quality synthesis. As a result, there has been significant interest in designing models that perform plug-and-play generation, i.e., to use a predefined or pretrained model, which is not explicitly trained on the generative task, to guide the generative process (e.g., using language). However, such guidance is typically useful only towards synthesizing high-level semantics rather than editing fine-grained details as in image-to-image translation tasks. To this end, and capitalizing on the powerful fine-grained generative control offered by the recent diffusion-based generative models, we introduce Steered Diffusion, a generalized framework for photorealistic zero-shot conditional image generation using a diffusion model trained for unconditional generation. The key idea is to steer the image generation of the diffusion model at inference time via designing a loss using a pre-trained inverse model that characterizes the conditional task. This loss modulates the sampling trajectory of the diffusion process. Our framework allows for easy incorporation of multiple conditions during inference. We present experiments using steered diffusion on several tasks including inpainting, colorization, text-guided semantic editing, and image super-resolution. Our results demonstrate clear qualitative and quantitative improvements over state-of-the-art diffusion-based plug-and-play models while adding negligible additional computational cost.

We propose Lizard, a linearization framework that transforms pretrained Transformer-based Large Language Models (LLMs) into flexible, subquadratic architectures for infinite-context generation. Transformer-based LLMs face significant memory and computational bottlenecks as context lengths increase, due to the quadratic complexity of softmax attention and the growing key-value (KV) cache. Lizard addresses these limitations by introducing a subquadratic attention mechanism that closely approximates softmax attention while preserving the output quality. Unlike previous linearization methods, which are often limited by fixed model structures and therefore exclude gating mechanisms, Lizard incorporates a gating module inspired by recent state-of-the-art linear models. This enables adaptive memory control, supports constant-memory inference, offers strong length generalization, and allows more flexible model design. Lizard combines gated linear attention for global context compression with sliding window attention enhanced by meta memory, forming a hybrid mechanism that captures both long-range dependencies and fine-grained local interactions. Moreover, we introduce a hardware-aware algorithm that accelerates the training speed of our models. Extensive experiments show that Lizard achieves near-lossless recovery of the teacher model's performance across standard language modeling tasks, while significantly outperforming previous linearization methods. On the 5-shot MMLU benchmark, Lizard improves over prior models by 18 points and shows significant improvements on associative recall tasks.

Gaze following and social gaze prediction are fundamental tasks providing insights into human communication behaviors, intent, and social interactions. Most previous approaches addressed these tasks separately, either by designing highly specialized social gaze models that do not generalize to other social gaze tasks or by considering social gaze inference as an ad-hoc post-processing of the gaze following task. Furthermore, the vast majority of gaze following approaches have proposed static models that can handle only one person at a time, therefore failing to take advantage of social interactions and temporal dynamics. In this paper, we address these limitations and introduce a novel framework to jointly predict the gaze target and social gaze label for all people in the scene. The framework comprises of: (i) a temporal, transformer-based architecture that, in addition to image tokens, handles person-specific tokens capturing the gaze information related to each individual; (ii) a new dataset, VSGaze, that unifies annotation types across multiple gaze following and social gaze datasets. We show that our model trained on VSGaze can address all tasks jointly, and achieves state-of-the-art results for multi-person gaze following and social gaze prediction.

Generative diffusion models, including Stable Diffusion and Midjourney, can generate visually appealing, diverse, and high-resolution images for various applications. These models are trained on billions of internet-sourced images, raising significant concerns about the potential unauthorized use of copyright-protected images. In this paper, we examine whether it is possible to determine if a specific image was used in the training set, a problem known in the cybersecurity community and referred to as a membership inference attack. Our focus is on Stable Diffusion, and we address the challenge of designing a fair evaluation framework to answer this membership question. We propose a methodology to establish a fair evaluation setup and apply it to Stable Diffusion, enabling potential extensions to other generative models. Utilizing this evaluation setup, we execute membership attacks (both known and newly introduced). Our research reveals that previously proposed evaluation setups do not provide a full understanding of the effectiveness of membership inference attacks. We conclude that the membership inference attack remains a significant challenge for large diffusion models (often deployed as black-box systems), indicating that related privacy and copyright issues will persist in the foreseeable future.

Spiking Neural Networks (SNNs) promise energy-efficient vision, but applying them to RGB visual tracking remains difficult: Existing SNN tracking frameworks either do not fully align with spike-driven computation or do not fully leverage neurons' spatiotemporal dynamics, leading to a trade-off between efficiency and accuracy. To address this, we introduce SpikeTrack, a spike-driven framework for energy-efficient RGB object tracking. SpikeTrack employs a novel asymmetric design that uses asymmetric timestep expansion and unidirectional information flow, harnessing spatiotemporal dynamics while cutting computation. To ensure effective unidirectional information transfer between branches, we design a memory-retrieval module inspired by neural inference mechanisms. This module recurrently queries a compact memory initialized by the template to retrieve target cues and sharpen target perception over time. Extensive experiments demonstrate that SpikeTrack achieves the state-of-the-art among SNN-based trackers and remains competitive with advanced ANN trackers. Notably, it surpasses TransT on LaSOT dataset while consuming only 1/26 of its energy. To our knowledge, SpikeTrack is the first spike-driven framework to make RGB tracking both accurate and energy efficient. The code and models are available at https://github.com/faicaiwawa/SpikeTrack.

Hallucination is a common problem for Large Vision-Language Models (LVLMs) with long generations which is difficult to eradicate. The generation with hallucinations is partially inconsistent with the image content. To mitigate hallucination, current studies either focus on the process of model inference or the results of model generation, but the solutions they design sometimes do not deal appropriately with various types of queries and the hallucinations of the generations about these queries. To accurately deal with various hallucinations, we present a unified framework, Dentist, for hallucination mitigation. The core step is to first classify the queries, then perform different processes of hallucination mitigation based on the classification result, just like a dentist first observes the teeth and then makes a plan. In a simple deployment, Dentist can classify queries as perception or reasoning and easily mitigate potential hallucinations in answers which has been demonstrated in our experiments. On MMbench, we achieve a 13.44%/10.2%/15.8% improvement in accuracy on Image Quality, a Coarse Perception visual question answering (VQA) task, over the baseline InstructBLIP/LLaVA/VisualGLM.

Reinforcement learning (RL) has become a pivotal technology in the post-training phase of large language models (LLMs). Traditional task-colocated RL frameworks suffer from significant scalability bottlenecks, while task-separated RL frameworks face challenges in complex dataflows and the corresponding resource idling and workload imbalance. Moreover, most existing frameworks are tightly coupled with LLM training or inference engines, making it difficult to support custom-designed engines. To address these challenges, we propose AsyncFlow, an asynchronous streaming RL framework for efficient post-training. Specifically, we introduce a distributed data storage and transfer module that provides a unified data management and fine-grained scheduling capability in a fully streamed manner. This architecture inherently facilitates automated pipeline overlapping among RL tasks and dynamic load balancing. Moreover, we propose a producer-consumer-based asynchronous workflow engineered to minimize computational idleness by strategically deferring parameter update process within staleness thresholds. Finally, the core capability of AsynFlow is architecturally decoupled from underlying training and inference engines and encapsulated by service-oriented user interfaces, offering a modular and customizable user experience. Extensive experiments demonstrate an average of 1.59 throughput improvement compared with state-of-the-art baseline. The presented architecture in this work provides actionable insights for next-generation RL training system designs.

Low-Rank Adaptation (LoRA) has become a widely adopted parameter-efficient fine-tuning (PEFT) technique for adapting large language models (LLMs) to downstream tasks. While prior work has explored strategies for integrating LLM training and serving, there still remains a gap in unifying fine-tuning and inference for LoRA-based models. We present Loquetier, a virtualized multi-LoRA framework that seamlessly integrates LoRA fine-tuning and serving within a single runtime. Loquetier introduces two key components: (1) a Virtualized Module that isolates PEFT-based modifications and supports multiple adapters on a shared base model, and (2) an optimized computation flow with a kernel design that merges fine-tuning and inference paths in forward propagation, enabling efficient batching and minimizing kernel invocation overhead. Extensive experiments across three task settings show that Loquetier consistently outperforms existing baselines in both performance and flexibility, achieving up to 3.0times the throughput of the state-of-the-art co-serving system on inference-only tasks and 46.4times higher SLO attainment than PEFT on unified fine-tuning and inference tasks. The implementation of Loquetier is publicly available at https://github.com/NJUDeepEngine/Loquetier.

Deep learning models are increasingly deployed on resource-constrained edge devices for real-time data analytics. In recent years, Vision Transformer models and their variants have demonstrated outstanding performance across various computer vision tasks. However, their high computational demands and inference latency pose significant challenges for model deployment on resource-constraint edge devices. To address this issue, we propose a novel Vision Transformer splitting framework, ED-ViT, designed to execute complex models across multiple edge devices efficiently. Specifically, we partition Vision Transformer models into several sub-models, where each sub-model is tailored to handle a specific subset of data classes. To further minimize computation overhead and inference latency, we introduce a class-wise pruning technique that reduces the size of each sub-model. We conduct extensive experiments on five datasets with three model structures, demonstrating that our approach significantly reduces inference latency on edge devices and achieves a model size reduction of up to 28.9 times and 34.1 times, respectively, while maintaining test accuracy comparable to the original Vision Transformer. Additionally, we compare ED-ViT with two state-of-the-art methods that deploy CNN and SNN models on edge devices, evaluating accuracy, inference time, and overall model size. Our comprehensive evaluation underscores the effectiveness of the proposed ED-ViT framework.

Human video generation is becoming an increasingly important task with broad applications in graphics, entertainment, and embodied AI. Despite the rapid progress of video diffusion models (VDMs), their use for general-purpose human video generation remains underexplored, with most works constrained to image-to-video setups or narrow domains like dance videos. In this work, we propose CAMEO, a cascaded framework for general human motion video generation. It seamlessly bridges Text-to-Motion (T2M) models and conditional VDMs, mitigating suboptimal factors that may arise in this process across both training and inference through carefully designed components. Specifically, we analyze and prepare both textual prompts and visual conditions to effectively train the VDM, ensuring robust alignment between motion descriptions, conditioning signals, and the generated videos. Furthermore, we introduce a camera-aware conditioning module that connects the two stages, automatically selecting viewpoints aligned with the input text to enhance coherence and reduce manual intervention. We demonstrate the effectiveness of our approach on both the MovieGen benchmark and a newly introduced benchmark tailored to the T2M-VDM combination, while highlighting its versatility across diverse use cases.

The Cambrian explosion of easily accessible pre-trained diffusion models suggests a demand for methods that combine multiple different pre-trained diffusion models without incurring the significant computational burden of re-training a larger combined model. In this paper, we cast the problem of combining multiple pre-trained diffusion models at the generation stage under a novel proposed framework termed superposition. Theoretically, we derive superposition from rigorous first principles stemming from the celebrated continuity equation and design two novel algorithms tailor-made for combining diffusion models in SuperDiff. SuperDiff leverages a new scalable It\^o density estimator for the log likelihood of the diffusion SDE which incurs no additional overhead compared to the well-known Hutchinson's estimator needed for divergence calculations. We demonstrate that SuperDiff is scalable to large pre-trained diffusion models as superposition is performed solely through composition during inference, and also enjoys painless implementation as it combines different pre-trained vector fields through an automated re-weighting scheme. Notably, we show that SuperDiff is efficient during inference time, and mimics traditional composition operators such as the logical OR and the logical AND. We empirically demonstrate the utility of using SuperDiff for generating more diverse images on CIFAR-10, more faithful prompt conditioned image editing using Stable Diffusion, and improved unconditional de novo structure design of proteins. https://github.com/necludov/super-diffusion

With recent advancements in natural language processing, Large Language Models (LLMs) have emerged as powerful tools for various real-world applications. Despite their prowess, the intrinsic generative abilities of LLMs may prove insufficient for handling complex tasks which necessitate a combination of task planning and the usage of external tools. In this paper, we first propose a structured framework tailored for LLM-based AI Agents and discuss the crucial capabilities necessary for tackling intricate problems. Within this framework, we design two distinct types of agents (i.e., one-step agent and sequential agent) to execute the inference process. Subsequently, we instantiate the framework using various LLMs and evaluate their Task Planning and Tool Usage (TPTU) abilities on typical tasks. By highlighting key findings and challenges, our goal is to provide a helpful resource for researchers and practitioners to leverage the power of LLMs in their AI applications. Our study emphasizes the substantial potential of these models, while also identifying areas that need more investigation and improvement.

Autoregressive video generation has improved rapidly in visual fidelity and interactivity, but it still suffers from long-term inconsistency and memory degradation. Most existing solutions either compress historical frames using predefined strategies or retrieve keyframes based on coarse implicit attention signals, both of which fail to handle evolving prompts with shifting entity references, leading to identity drift, character duplication, and attribute loss. To address this, we propose IAMFlow, a training-free identity-aware memory framework that explicitly models and tracks persistent entity identities, enabling consistent generation across prompt transitions. Specifically, an LLM extracts entities with visual attributes from each prompt and assigns unique global IDs for identity-aware memory, while a VLM asynchronously verifies and refines attributes from rendered frames, enabling explicit entity tracking in place of implicit similarity-based matching. To keep the proposed framework computationally practical, we design a systematic inference acceleration pipeline, including asynchronous visual verification, adaptive prompt transition, and model quantization, which achieves faster generation than existing baselines. Furthermore, we introduce NarraStream-Bench, a benchmark for narrative streaming video generation that features 324 multi-prompt scripts spanning six dimensions and a three-dimensional evaluation protocol that integrates both traditional metrics and multimodal large language model-based assessments. Extensive experiments show that IAMFlow, despite being training-free, achieves the best overall performance on NarraStream-Bench, outperforming the strongest baseline by 2.56 points, while achieving a 1.39times speedup over the most efficient baseline in the 60-second multi-prompt setting.

The widespread popularity of Large Language Models (LLMs), partly due to their unique ability to perform in-context learning, has also brought to light the importance of ethical and safety considerations when deploying these pre-trained models. In this work, we focus on investigating machine unlearning for LLMs motivated by data protection regulations. In contrast to the growing literature on fine-tuning methods to achieve unlearning, we focus on a comparatively lightweight alternative called soft prompting to realize the unlearning of a subset of training data. With losses designed to enforce forgetting as well as utility preservation, our framework Soft Prompting for Unlearning (SPUL) learns prompt tokens that can be appended to an arbitrary query to induce unlearning of specific examples at inference time without updating LLM parameters. We conduct a rigorous evaluation of the proposed method and our results indicate that SPUL can significantly improve the trade-off between utility and forgetting in the context of text classification and question answering with LLMs. We further validate our method using multiple LLMs to highlight the scalability of our framework and provide detailed insights into the choice of hyperparameters and the influence of the size of unlearning data. Our implementation is available at https://github.com/karuna-bhaila/llm_unlearning.

As LLM-based multi-agent systems (MAS) become more autonomous, their free-form interactions increasingly dominate system behavior. However, scaling the number of agents often amplifies context pressure, coordination errors, and system drift. It is well known that building robust MAS requires more than prompt tuning or increased model intelligence. It necessitates engineering discipline focused on architecture to manage complexity under uncertainty. We characterize agentic software by a core property: runtime generation and evolution under uncertainty. Drawing upon and extending software engineering experience, especially object-oriented programming, this paper introduces Loosely-Structured Software (LSS), a new class of software systems that shifts the engineering focus from constructing deterministic logic to managing the runtime entropy generated by View-constructed programming, semantic-driven self-organization, and endogenous evolution. To make this entropy governable, we introduce design principles under a three-layer engineering framework: View/Context Engineering to manage the execution environment and maintain task-relevant Views, Structure Engineering to organize dynamic binding over artifacts and agents, and Evolution Engineering to govern the lifecycle of self-rewriting artifacts. Building on this framework, we develop LSS design patterns as semantic control blocks that stabilize fluid, inference-mediated interactions while preserving agent adaptability. Together, these abstractions improve the designability, scalability, and evolvability of agentic infrastructure. We provide basic experimental validation of key mechanisms, demonstrating the effectiveness of LSS.

This paper introduces a parallel and asynchronous Transformer framework designed for efficient and accurate multilingual lip synchronization in real-time video conferencing systems. The proposed architecture integrates translation, speech processing, and lip-synchronization modules within a pipeline-parallel design that enables concurrent module execution through message-queue-based decoupling, reducing end-to-end latency by up to 3.1 times compared to sequential approaches. To enhance computational efficiency and throughput, the inference workflow of each module is optimized through low-level graph compilation, mixed-precision quantization, and hardware-accelerated kernel fusion. These optimizations provide substantial gains in efficiency while preserving model accuracy and visual quality. In addition, a context-adaptive silence-detection component segments the input speech stream at semantically coherent boundaries, improving translation consistency and temporal alignment across languages. Experimental results demonstrate that the proposed parallel architecture outperforms conventional sequential pipelines in processing speed, synchronization stability, and resource utilization. The modular, message-oriented design makes this work applicable to resource-constrained IoT communication scenarios including telemedicine, multilingual kiosks, and remote assistance systems. Overall, this work advances the development of low-latency, resource-efficient multimodal communication frameworks for next-generation AIoT systems.

Recent Large Language Models have been enhanced with vision capabilities, enabling them to comprehend images, videos, and interleaved vision-language content. However, the learning methods of these large multimodal models typically treat videos as predetermined clips, making them less effective and efficient at handling streaming video inputs. In this paper, we propose a novel Learning-In-Video-Stream (LIVE) framework, which enables temporally aligned, long-context, and real-time conversation within a continuous video stream. Our LIVE framework comprises comprehensive approaches to achieve video streaming dialogue, encompassing: (1) a training objective designed to perform language modeling for continuous streaming inputs, (2) a data generation scheme that converts offline temporal annotations into a streaming dialogue format, and (3) an optimized inference pipeline to speed up the model responses in real-world video streams. With our LIVE framework, we built VideoLLM-online model upon Llama-2/Llama-3 and demonstrate its significant advantages in processing streaming videos. For instance, on average, our model can support streaming dialogue in a 5-minute video clip at over 10 FPS on an A100 GPU. Moreover, it also showcases state-of-the-art performance on public offline video benchmarks, such as recognition, captioning, and forecasting. The code, model, data, and demo have been made available at https://showlab.github.io/videollm-online.

LLMs have transformed NLP, yet deploying them on edge devices poses great carbon challenges. Prior estimators remain incomplete, neglecting peripheral energy use, distinct prefill/decode behaviors, and SoC design complexity. This paper presents CO2-Meter, a unified framework for estimating operational and embodied carbon in LLM edge inference. Contributions include: (1) equation-based peripheral energy models and datasets; (2) a GNN-based predictor with phase-specific LLM energy data; (3) a unit-level embodied carbon model for SoC bottleneck analysis; and (4) validation showing superior accuracy over prior methods. Case studies show CO2-Meter's effectiveness in identifying carbon hotspots and guiding sustainable LLM design on edge platforms. Source code: https://github.com/fuzhenxiao/CO2-Meter

Multimodal large language models (MLLMs) exhibit impressive capabilities but still face challenges in complex visual reasoning. While recent efforts attempt to enhance MLLMs' reasoning by incorporating OpenAI o1-like structured thinking through explicit search structures or teacher-guided distillation, they often struggle to balance performance and efficiency. A critical limitation is their heavy reliance on extensive data and search spaces, resulting in low-efficiency implicit insight extraction and data utilization. To address this, we propose AStar, an Automated Structured thinking paradigm for multimodal reasoning via Monte Carlo Tree Search (MCTS). AStar automatically derives high-level cognitive reasoning patterns from limited data using MCTS-powered hierarchical structures. Building on these explicit patterns, we design a unified reasoning framework that seamlessly integrates models' internal reasoning capabilities and external reasoning guidelines, enabling efficient inference with minimal tree iterations. This novel paradigm strikes a compelling balance between performance and efficiency. Extensive experiments demonstrate AStar's effectiveness, achieving superior accuracy (54.0%) on the MathVerse benchmark with a 7B backbone, surpassing GPT-4o (50.2%) while maintaining substantial data and computational efficiency.

Large language models are typically adapted to downstream tasks through supervised fine-tuning on domain-specific data. While standard fine-tuning focuses on minimizing generation loss to optimize model parameters, we take a deeper step by retaining and leveraging the model's own learning signals, analogous to how human learners reflect on past mistakes to improve future performance. We first introduce the concept of Mistake Log to systematically track the model's learning behavior and recurring errors throughout fine-tuning. Treating the original transformer-based model as the Pilot, we correspondingly design a Copilot model to refine the Pilot's inference performance via logits rectification. We name the overall Pilot-Copilot framework the Transformer Copilot, which introduces (i) a novel Copilot model design, (ii) a joint training paradigm where the Copilot continuously learns from the evolving Mistake Log alongside the Pilot, and (iii) a fused inference paradigm where the Copilot rectifies the Pilot's logits for enhanced generation. We provide both theoretical and empirical analyses on our new learning framework. Experiments on 12 benchmarks spanning commonsense, arithmetic, and recommendation tasks demonstrate that Transformer Copilot consistently improves performance by up to 34.5%, while introducing marginal computational overhead to Pilot models and exhibiting strong scalability and transferability.

Scaling inference-time computation has enabled Large Language Models (LLMs) to achieve strong reasoning performance, but inherently sequential decoding leads to substantial latency, especially on complex tasks. Recent work on adaptive parallel reasoning aims to improve inference efficiency by decomposing the problem-solving process into concurrent reasoning threads when beneficial. However, existing methods on realistic tasks are either limited to supervised behavior cloning or exhibit significant accuracy drops compared to widely-used sequential long chain-of-thought (CoT) baselines. Moreover, many require customized inference engines, complicating deployment. We introduce ThreadWeaver, a framework for adaptive parallel reasoning that achieves accuracy on par with popular sequential reasoning models of comparable size while significantly reducing inference latency. ThreadWeaver's performance stems from three key innovations: 1) a two-stage parallel trajectory generator that produces large-scale, high-quality CoT data with parallel annotations for supervised fine-tuning; 2) a trie-based training-inference co-design that enables parallel reasoning on any off-the-shelf autoregressive inference engine without modifying position embeddings or KV caches; and 3) a parallelization-aware reinforcement learning framework that teaches the model to balance accuracy with effective parallelization. Across six challenging mathematical reasoning benchmarks, ThreadWeaver trained atop Qwen3-8B achieves accuracy comparable to cutting-edge sequential reasoning models (71.9% on average and 79.9% on AIME24) while delivering up to 1.53x average speedup in token latency, establishing a new Pareto frontier between accuracy and efficiency.

Recent advances in stereo matching have achieved remarkable accuracy, but often rely on large models, heavy computation, or additional foundation-model priors, making them difficult to deploy on resource-constrained platforms. In contrast, efficient stereo models offer faster inference but are commonly considered less capable of strong zero-shot generalization. In this paper, we challenge this assumption by introducing Lite Any Stereo V2 (LAS2), an ultra-fast model series designed for efficient zero-shot stereo matching. LAS2 is developed from both architecture and training perspectives. Architecturally, we revisit efficient stereo design under practical deployment settings and propose a 2D-only cost aggregation framework, optimized for real inference latency rather than theoretical MACs alone. For training, we develop a three-stage strategy that combines synthetic supervision, self-distillation, and real-world knowledge distillation. To improve the reliability of real-world pseudo supervision, we further introduce pseudo-label filtering and an error-clamping operation, enabling smoother synthetic-to-real transfer. We instantiate LAS2 as a family of models, including feed-forward variants for different efficiency budgets and an iterative variant for higher accuracy. Extensive experiments show that LAS2 achieves state-of-the-art accuracy among efficient stereo methods while maintaining significantly lower latency. Specifically, LAS2-H achieves stronger overall zero-shot performance than the iterative method Fast-FoundationStereo, with 1.8x and 2.7x faster inference on H200 and Orin, respectively. The project page, demos, and code are available at https://tomtomtommi.github.io/LiteAnyStereoV2/.

We present a direct speech-to-speech translation (S2ST) model that translates speech from one language to speech in another language without relying on intermediate text generation. We tackle the problem by first applying a self-supervised discrete speech encoder on the target speech and then training a sequence-to-sequence speech-to-unit translation (S2UT) model to predict the discrete representations of the target speech. When target text transcripts are available, we design a joint speech and text training framework that enables the model to generate dual modality output (speech and text) simultaneously in the same inference pass. Experiments on the Fisher Spanish-English dataset show that the proposed framework yields improvement of 6.7 BLEU compared with a baseline direct S2ST model that predicts spectrogram features. When trained without any text transcripts, our model performance is comparable to models that predict spectrograms and are trained with text supervision, showing the potential of our system for translation between unwritten languages. Audio samples are available at https://facebookresearch.github.io/speech_translation/direct_s2st_units/index.html .