Large language models (LLMs) achieve impressive performance across domains but face significant challenges when deployed on consumer-grade GPUs or personal devices such as laptops, due to high memory consumption and inference costs. Post-training quantization (PTQ) of LLMs offers a promising solution that reduces their memory footprint and decoding latency. In practice, PTQ with uniform quantization representation is favored for its efficiency and ease of deployment since uniform quantization is widely supported by mainstream hardware and software libraries. Recent studies on $\geq 2$-bit uniform quantization have led to noticeable improvements in post-quantization model performance; however, they primarily focus on quantization methodologies, while the initialization of quantization parameters is underexplored and still relies on the suboptimal Min-Max strategies. In this work, we propose NeUQI, a method devoted to efficiently determining near-optimal initial parameters for uniform quantization. NeUQI is orthogonal to prior quantization methodologies and can seamlessly integrate with them. The experiments with the LLaMA and Qwen families on various tasks demonstrate that our NeUQI consistently outperforms existing methods. Furthermore, when combined with a lightweight distillation strategy, NeUQI can achieve superior performance to PV-tuning, a much more resource-intensive approach.
Vision-Language-Action (VLA) models have shown impressive capabilities across a wide range of robotics manipulation tasks. However, their growing model size poses significant challenges for deployment on resource-constrained robotic systems. While 1-bit pretraining has proven effective for enhancing the inference efficiency of large language models with minimal performance loss, its application to VLA models remains underexplored. In this work, we present BitVLA, the first 1-bit VLA model for robotics manipulation, in which every parameter is ternary, i.e., {-1, 0, 1}. To further reduce the memory footprint of the vision encoder, we propose the distillation-aware training strategy that compresses the full-precision encoder to 1.58-bit weights. During this process, a full-precision encoder serves as a teacher model to better align latent representations. Despite the lack of large-scale robotics pretraining, BitVLA achieves performance comparable to the state-of-the-art model OpenVLA-OFT with 4-bit post-training quantization on the LIBERO benchmark, while consuming only 29.8% of the memory. These results highlight BitVLA's promise for deployment on memory-constrained edge devices. We release the code and model weights in https://github.com/ustcwhy/BitVLA.
Recent advancements in large language models (LLMs) have revitalized philosophical debates surrounding artificial intelligence. Two of the most fundamental challenges - namely, the Frame Problem and the Symbol Grounding Problem - have historically been viewed as unsolvable within traditional symbolic AI systems. This study investigates whether modern LLMs possess the cognitive capacities required to address these problems. To do so, I designed two benchmark tasks reflecting the philosophical core of each problem, administered them under zero-shot conditions to 13 prominent LLMs (both closed and open-source), and assessed the quality of the models' outputs across five trials each. Responses were scored along multiple criteria, including contextual reasoning, semantic coherence, and information filtering. The results demonstrate that while open-source models showed variability in performance due to differences in model size, quantization, and instruction tuning, several closed models consistently achieved high scores. These findings suggest that select modern LLMs may be acquiring capacities sufficient to produce meaningful and stable responses to these long-standing theoretical challenges.
This work addresses a fundamental barrier in recommender systems: the inability to generalize across domains without extensive retraining. Traditional ID-based approaches fail entirely in cold-start and cross-domain scenarios where new users or items lack sufficient interaction history. Inspired by foundation models' cross-domain success, we develop a foundation model for sequential recommendation that achieves genuine zero-shot generalization capabilities. Our approach fundamentally departs from existing ID-based methods by deriving item representations exclusively from textual features. This enables immediate embedding of any new item without model retraining. We introduce unified item tokenization with Finite Scalar Quantization that transforms heterogeneous textual descriptions into standardized discrete tokens. This eliminates domain barriers that plague existing systems. Additionally, the framework features hybrid bidirectional-causal attention that captures both intra-item token coherence and inter-item sequential dependencies. An efficient catalog-aware beam search decoder enables real-time token-to-item mapping. Unlike conventional approaches confined to their training domains, RecGPT naturally bridges diverse recommendation contexts through its domain-invariant tokenization mechanism. Comprehensive evaluations across six datasets and industrial scenarios demonstrate consistent performance advantages.
Monocular Depth Estimation (MDE) has emerged as a pivotal task in computer vision, supporting numerous real-world applications. However, deploying accurate depth estimation models on resource-limited edge devices, especially Application-Specific Integrated Circuits (ASICs), is challenging due to the high computational and memory demands. Recent advancements in foundational depth estimation deliver impressive results but further amplify the difficulty of deployment on ASICs. To address this, we propose QuartDepth which adopts post-training quantization to quantize MDE models with hardware accelerations for ASICs. Our approach involves quantizing both weights and activations to 4-bit precision, reducing the model size and computation cost. To mitigate the performance degradation, we introduce activation polishing and compensation algorithm applied before and after activation quantization, as well as a weight reconstruction method for minimizing errors in weight quantization. Furthermore, we design a flexible and programmable hardware accelerator by supporting kernel fusion and customized instruction programmability, enhancing throughput and efficiency. Experimental results demonstrate that our framework achieves competitive accuracy while enabling fast inference and higher energy efficiency on ASICs, bridging the gap between high-performance depth estimation and practical edge-device applicability. Code: https://github.com/shawnricecake/quart-depth
Commonly used image tokenizers produce a 2D grid of spatially arranged tokens. In contrast, so-called 1D image tokenizers represent images as highly compressed one-dimensional sequences of as few as 32 discrete tokens. We find that the high degree of compression achieved by a 1D tokenizer with vector quantization enables image editing and generative capabilities through heuristic manipulation of tokens, demonstrating that even very crude manipulations -- such as copying and replacing tokens between latent representations of images -- enable fine-grained image editing by transferring appearance and semantic attributes. Motivated by the expressivity of the 1D tokenizer's latent space, we construct an image generation pipeline leveraging gradient-based test-time optimization of tokens with plug-and-play loss functions such as reconstruction or CLIP similarity. Our approach is demonstrated for inpainting and text-guided image editing use cases, and can generate diverse and realistic samples without requiring training of any generative model.
Transforming complex actions into discrete skill abstractions has demonstrated strong potential for robotic manipulation. Existing approaches mainly leverage latent variable models, e.g., VQ-VAE, to learn skill abstractions through learned vectors (codebooks), while they suffer from codebook collapse and modeling the causal relationship between learned skills. To address these limitations, we present \textbf{S}kill \textbf{T}raining with \textbf{A}ugmented \textbf{R}otation (\textbf{STAR}), a framework that advances both skill learning and composition to complete complex behaviors. Specifically, to prevent codebook collapse, we devise rotation-augmented residual skill quantization (RaRSQ). It encodes relative angles between encoder outputs into the gradient flow by rotation-based gradient mechanism. Points within the same skill code are forced to be either pushed apart or pulled closer together depending on gradient directions. Further, to capture the causal relationship between skills, we present causal skill transformer (CST) which explicitly models dependencies between skill representations through an autoregressive mechanism for coherent action generation. Extensive experiments demonstrate the superiority of STAR on both LIBERO benchmark and realworld tasks, with around 12\% improvement over the baselines.
As large language models (LLMs) are increasingly used in legal applications, current evaluation benchmarks tend to focus mainly on factual accuracy while largely neglecting important linguistic quality aspects such as clarity, coherence, and terminology. To address this gap, we propose three steps: First, we develop a regression model to evaluate the quality of legal texts based on clarity, coherence, and terminology. Second, we create a specialized set of legal questions. Third, we analyze 49 LLMs using this evaluation framework. Our analysis identifies three key findings: First, model quality levels off at 14 billion parameters, with only a marginal improvement of $2.7\%$ noted at 72 billion parameters. Second, engineering choices such as quantization and context length have a negligible impact, as indicated by statistical significance thresholds above 0.016. Third, reasoning models consistently outperform base architectures. A significant outcome of our research is the release of a ranking list and Pareto analysis, which highlight the Qwen3 series as the optimal choice for cost-performance tradeoffs. This work not only establishes standardized evaluation protocols for legal LLMs but also uncovers fundamental limitations in current training data refinement approaches. Code and models are available at: https://github.com/lyxx3rd/LegalEval-Q.
With the increasing size of frontier LLMs, post-training quantization has become the standard for memory-efficient deployment. Recent work has shown that basic rounding-based quantization schemes pose security risks, as they can be exploited to inject malicious behaviors into quantized models that remain hidden in full precision. However, existing attacks cannot be applied to more complex quantization methods, such as the GGUF family used in the popular ollama and llama$.$cpp frameworks. In this work, we address this gap by introducing the first attack on GGUF. Our key insight is that the quantization error -- the difference between the full-precision weights and their (de-)quantized version -- provides sufficient flexibility to construct malicious quantized models that appear benign in full precision. Leveraging this, we develop an attack that trains the target malicious LLM while constraining its weights based on quantization errors. We demonstrate the effectiveness of our attack on three popular LLMs across nine GGUF quantization data types on three diverse attack scenarios: insecure code generation ($\Delta$=$88.7\%$), targeted content injection ($\Delta$=$85.0\%$), and benign instruction refusal ($\Delta$=$30.1\%$). Our attack highlights that (1) the most widely used post-training quantization method is susceptible to adversarial interferences, and (2) the complexity of quantization schemes alone is insufficient as a defense.
Climate Finance Bench introduces an open benchmark that targets question-answering over corporate climate disclosures using Large Language Models. We curate 33 recent sustainability reports in English drawn from companies across all 11 GICS sectors and annotate 330 expert-validated question-answer pairs that span pure extraction, numerical reasoning, and logical reasoning. Building on this dataset, we propose a comparison of RAG (retrieval-augmented generation) approaches. We show that the retriever's ability to locate passages that actually contain the answer is the chief performance bottleneck. We further argue for transparent carbon reporting in AI-for-climate applications, highlighting advantages of techniques such as Weight Quantization.
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 2.78$\times$ speedup across various tasks for the 4-bit weight Llama-3-70B model on an A100 GPU, outperforming EAGLE-2 by 1.31$\times$. Code available at https://github.com/AI9Stars/SpecMQuant.
The main goal of post-training quantization (PTQ) is to produced a compressed model whose output distribution is as close to the original model's as possible. To do this tractably, almost all LLM PTQ algorithms quantize linear layers by independently minimizing the immediate activation error. However, this localized objective ignores the effect of subsequent layers, so reducing it does not necessarily give a closer model. In this work, we introduce Yet Another Quantization Algorithm (YAQA), an adaptive rounding algorithm that uses Kronecker-factored approximations of each linear layer's Hessian with respect to the \textit{full model} KL divergence. YAQA consists of two components: Kronecker-factored sketches of the full layerwise Hessian that can be tractably computed for hundred-billion parameter LLMs, and a quantizer-independent rounding algorithm that uses these sketches and comes with theoretical guarantees. Across a wide range of models and quantizers, YAQA empirically reduces the KL divergence to the original model by $\approx 30\%$ while achieving state of the art performance on downstream tasks.
Model merging has emerged as a powerful technique for combining task-specific weights, achieving superior performance in multi-target domain adaptation. However, when applied to practical scenarios, such as quantized models, new challenges arise. In practical scenarios, quantization is often applied to target-specific data, but this process restricts the domain of interest and introduces discretization effects, making model merging highly non-trivial. In this study, we analyze the impact of quantization on model merging through the lens of error barriers. Leveraging these insights, we propose a novel post-training quantization, HDRQ - Hessian and distant regularizing quantization - that is designed to consider model merging for multi-target domain adaptation. Our approach ensures that the quantization process incurs minimal deviation from the source pre-trained model while flattening the loss surface to facilitate smooth model merging. To our knowledge, this is the first study on this challenge, and extensive experiments confirm its effectiveness.
Generative models have excelled in audio tasks using approaches such as language models, diffusion, and flow matching. However, existing generative approaches for speech enhancement (SE) face notable challenges: language model-based methods suffer from quantization loss, leading to compromised speaker similarity and intelligibility, while diffusion models require complex training and high inference latency. To address these challenges, we propose FlowSE, a flow-matching-based model for SE. Flow matching learns a continuous transformation between noisy and clean speech distributions in a single pass, significantly reducing inference latency while maintaining high-quality reconstruction. Specifically, FlowSE trains on noisy mel spectrograms and optional character sequences, optimizing a conditional flow matching loss with ground-truth mel spectrograms as supervision. It implicitly learns speech's temporal-spectral structure and text-speech alignment. During inference, FlowSE can operate with or without textual information, achieving impressive results in both scenarios, with further improvements when transcripts are available. Extensive experiments demonstrate that FlowSE significantly outperforms state-of-the-art generative methods, establishing a new paradigm for generative-based SE and demonstrating the potential of flow matching to advance the field. Our code, pre-trained checkpoints, and audio samples are available.
The efficiency of attention is critical because its time complexity grows quadratically with sequence length. SageAttention2 addresses this by utilizing quantization to accelerate matrix multiplications (Matmul) in attention. To further accelerate SageAttention2, we propose to utilize the faster instruction of FP8 Matmul accumulated in FP16. The instruction is 2x faster than the FP8 Matmul used in SageAttention2. Our experiments show that SageAttention2++ achieves a 3.9x speedup over FlashAttention while maintaining the same attention accuracy as SageAttention2. This means SageAttention2++ effectively accelerates various models, including those for language, image, and video generation, with negligible end-to-end metrics loss. The code will be available at https://github.com/thu-ml/SageAttention.
Selecting artificial intelligence (AI) models, such as large language models (LLMs), from multiple candidates requires accurate performance estimation. This is ideally achieved through empirical evaluations involving abundant real-world data. However, such evaluations are costly and impractical at scale. To address this challenge, autoevaluation methods leverage synthetic data produced by automated evaluators, such as LLMs-as-judges, reducing variance but potentially introducing bias. Recent approaches have employed semi-supervised prediction-powered inference (\texttt{PPI}) to correct for the bias of autoevaluators. However, the use of autoevaluators may lead in practice to a degradation in sample efficiency compared to conventional methods using only real-world data. In this paper, we propose \texttt{R-AutoEval+}, a novel framework that provides finite-sample reliability guarantees on the model evaluation, while also ensuring an enhanced (or at least no worse) sample efficiency compared to conventional methods. The key innovation of \texttt{R-AutoEval+} is an adaptive construction of the model evaluation variable, which dynamically tunes its reliance on synthetic data, reverting to conventional methods when the autoevaluator is insufficiently accurate. Experiments on the use of LLMs-as-judges for the optimization of quantization settings for the weights of an LLM, and for prompt design in LLMs confirm the reliability and efficiency of \texttt{R-AutoEval+}.
Post-training compression reduces the computational and memory costs of large language models (LLMs), enabling resource-efficient deployment. However, existing compression benchmarks only focus on language modeling (e.g., perplexity) and natural language understanding tasks (e.g., GLUE accuracy), ignoring the agentic capabilities - workflow, tool use/function call, long-context understanding and real-world application. We introduce the Agent Compression Benchmark (ACBench), the first comprehensive benchmark for evaluating how compression impacts LLMs' agentic abilities. ACBench spans (1) 12 tasks across 4 capabilities (e.g., WorfBench for workflow generation, Needle-in-Haystack for long-context retrieval), (2) quantization (GPTQ, AWQ) and pruning (Wanda, SparseGPT), and (3) 15 models, including small (Gemma-2B), standard (Qwen2.5 7B-32B), and distilled reasoning LLMs (DeepSeek-R1-Distill). Our experiments reveal compression tradeoffs: 4-bit quantization preserves workflow generation and tool use (1%-3% drop) but degrades real-world application accuracy by 10%-15%. We introduce ERank, Top-k Ranking Correlation and Energy to systematize analysis. ACBench provides actionable insights for optimizing LLM compression in agentic scenarios. The code can be found in https://github.com/pprp/ACBench.
Transformer models power many AI applications but suffer from high inference latency, limiting their use in real-time settings. Multi-device inference can reduce latency by parallelizing computation. Yet, existing methods require high inter-device bandwidth, making them impractical for bandwidth-constrained environments. We propose ASTRA, a communication-efficient framework that accelerates Transformer inference through a novel integration of sequence parallelism and a Mixed-Precision Attention mechanism designed to minimize inter-device communication. ASTRA compresses non-local token embeddings via vector quantization and preserves task accuracy through two optimizations, Noise-Augmented Quantization and Distributed Class Tokens. Experiments on ViT and GPT2 across vision and NLP tasks show that ASTRA achieves up to 2.64X speedups over single-device inference and up to 15.25X speedups over state-of-the-art multi-device inferences, while operating under bandwidths as low as 10 Mbps. ASTRA is open-sourced at https://github.com/xl1990/Astra.
We demonstrate, for the first time, fully quantized training (FQT) of large language models (LLMs) using predominantly 4-bit floating-point (FP4) precision for weights, activations, and gradients on datasets up to 200 billion tokens. We extensively investigate key design choices for FP4, including block sizes, scaling formats, and rounding methods. Our analysis shows that the NVFP4 format, where each block of 16 FP4 values (E2M1) shares a scale represented in E4M3, provides optimal results. We use stochastic rounding for backward and update passes and round-to-nearest for the forward pass to enhance stability. Additionally, we identify a theoretical and empirical threshold for effective quantized training: when the gradient norm falls below approximately $\sqrt{3}$ times the quantization noise, quantized training becomes less effective. Leveraging these insights, we successfully train a 7-billion-parameter model on 256 Intel Gaudi2 accelerators. The resulting FP4-trained model achieves downstream task performance comparable to a standard BF16 baseline, confirming that FP4 training is a practical and highly efficient approach for large-scale LLM training. A reference implementation is supplied in https://github.com/Anonymous1252022/fp4-all-the-way .
The ever-growing size of neural networks poses serious challenges on resource-constrained devices, such as embedded sensors. Compression algorithms that reduce their size can mitigate these problems, provided that model performance stays close to the original. We propose a novel post-training compression framework that combines rate-aware quantization with entropy coding by (1) extending the well-known layer-wise loss by a quadratic rate estimation, and (2) providing locally exact solutions to this modified objective following the Optimal Brain Surgeon (OBS) method. Our method allows for very fast decoding and is compatible with arbitrary quantization grids. We verify our results empirically by testing on various computer-vision networks, achieving a 20-40\% decrease in bit rate at the same performance as the popular compression algorithm NNCodec. Our code is available at https://github.com/Conzel/cerwu.