Quantization
-
Predicate Neuron: Differentiable Quantization of Predicate Relations in Gradient-based Learning
2026-10-27 12:07:22Ioannis Konstantoulas, Iliana Loi, Kyriakos SgarbasAbstract
In machine learning, symbolic reasoning approaches often rely on static architectures that hinder integration into deep neural networks and gradient-based optimization. To address this, we introduce the Native Predicate Neuron, a differentiable logic-reasoning neural module that enriches standard architectures with symbolic reasoning capabilities via first-order logic quantization. The proposed neuron models relational logic by learning to alternate between operations using a single trainable parameter. It features a tailored gradient mechanism with a dual-nature formulation: executing hard logical operations during training while maintaining differentiability for backpropagation. By quantizing both its input and choice state, the Predicate Neuron approximates continuous and non-continuous functions, effectively resolving discontinuities that standard deep learning architectures struggle with. We evaluate the Predicate Neuron across Fully-Connected, Convolutional, Recurrent, and Transformer architectures on benchmarks like MNIST and Wikitext. Results show that incorporating Predicate Neurons accelerates convergence, boosts accuracy, and lowers loss in general, with particularly strong improvements in tasks requiring monotonic reasoning or handling discontinuities. Overall, the Predicate Neuron natively integrates into standard neural networks to enable fuzzy logic operations and is available in Keras and PyTorch.
Abstract (translated)
URL
https://arxiv.org/abs/2610.37727
PDF
-
Quantized Edge Inference of a Self-Supervised SAR-Optical Encoder on Commodity x86 Hardware
2026-10-27 11:24:25Christian Sanchez-Leon, Dongming ChenAbstract
Self-supervised vision transformers are increasingly used as reusable encoders for Earth observation, but their inference cost on the hardware available in the field is rarely characterized. We deploy a pretrained SAR-optical masked-autoencoder encoder on three commodity x86 nodes, from a low-power CPU without AVX2 to an AVX2+VNNI reference, at FP32, FP16 and INT8 on the CPU and, where present, the integrated GPU, and report latency, energy (RAPL and wall plug) and linear-probe accuracy. Quantization is accuracy-neutral: INT8 changes micro-mAP by only +0.0006, within the seed spread, so precision becomes a pure cost decision. INT8 lowers CPU latency on every node, by 1.45x even on the CPU without AVX2, and CPU energy wherever it is measured, and the VNNI reference (1.53x) suggests a further gain beyond AVX2. The Pareto-optimal device shifts from the integrated GPU, which is 4.38x faster than the CPU on the weakest node, to the CPU as the CPU class improves.
Abstract (translated)
URL
https://arxiv.org/abs/2610.37056
PDF
-
AxisQ: Depth-Adaptive Channel--Frequency Quantization for Camouflaged Object Detection
2026-10-27 08:41:01Chi Ruan, Tianqi Li, Xin He, Xue Geng, Ming-Ming Cheng, Yun LiuAbstract
Camouflaged object detection (COD) depends on weak texture and boundary cues that can be destroyed by aggressive post-training quantization (PTQ). Existing PTQ methods mainly optimize quantizers or reconstruct network blocks, while using a single precision-allocation domain throughout hierarchical backbones. Instead, we introduce AxisQ, a depth-adaptive bit-allocation framework motivated by the observation that the useful allocation axis changes with feature depth. Under a fixed per-layer average activation-bit budget, AxisQ allocates precision in the channel domain for shallow features and in the local-frequency domain for deep features, using COD-aware sensitivity to better preserve subtle foreground regions and ambiguous boundaries. We also formulate a boundary-aware Hessian approximation for reconstruction by incorporating boundary-weighted gradient and error statistics. Experiments across multiple COD models and diverse benchmarks demonstrate improved accuracy over the strongest baselines at low bit widths, with relative weighted F-measure gains of 12.7% for CFRN and 33.3% for CamoFormer-P on CAMO under W2A2 quantization. These results support adapting the allocation domain to feature depth as a transferable quantization strategy for hierarchical COD models.
Abstract (translated)
URL
https://arxiv.org/abs/2610.36844
PDF
-
Tri-Directional Sub-Voxel Displacement and Ipsilateral Territory Decoupling for 3D Bilateral Coronary Artery Tree Reconstruction in Cardiac CT Angiography
2026-10-27 02:55:53Zeyu Jia, Xin MingAbstract
Three-dimensional (3D) reconstruction of bilateral coronary trees (LCA and RCA) from cardiac computed tomography angiography (CCTA) is limited by $64^3$ quantization, territory ambiguity, and sinotubular junction (STJ) ostium drift. We present a canonical $SO(3)$ $(Z_{\text{IS}}, Y_{\text{PA}}, X_{\text{LR}})$ multi-scale representation and Riemannian arborescence framework. First, we map 3D CCTA volumes, centerlines, and thick-stem ostia into $(Z_{\text{IS}}, Y_{\text{PA}}, X_{\text{LR}})$ space, extending a 3D pyramid and tubular decoder with a $128^3 \to 64^3$ boundary skip, a stop-gradient $128^3$ residual branch, a sub-voxel displacement head $\Delta\hat{\mathbf{u}}(\mathbf{x})$, a dual-channel ipsilateral head $(\mathbf{P}_{\text{LCA}}, \mathbf{P}_{\text{RCA}})$, and a $32^3$ ostium detector. Second, all $327,385$ active parameters ($398,843$ total) are trained in one 30-epoch AdamW loop ($\epsilon = 10^{-8}$) under decoupled L2-SP and aortic-wall consistency. Third, our Riemannian solver couples cylindrical Valsalva-STJ calibration, $128^3$ orifice contact rooting, contralateral inhibition, centripetal pruning, and 4-stage curvature-adaptive normal-plane pullback. On 14 unseen ImageCAS development volumes (28 trees; cases 201--250 isolated), the framework achieves $2.1652 \pm 0.4686\text{ mm}$ $\text{CD}_{\text{CL}}$ ($\text{CD}_{\text{CL}}^{\text{arc1mm}} = 2.2697 \pm 0.4863\text{ mm}$, $p \le 0.00024$; $57.40\% / 78.20\%$ $\text{F1}@1/2\text{mm}$; $4.6880 / 4.9298\text{ mm}$ ostia). Zero-shot evaluation on 50 unseen volumes (Cases 251--300, 100 trees) yields $2.9389 \pm 1.2785\text{ mm}$ $\text{CD}_{\text{CL}}$ ($p \le 7.24 \times 10^{-7}$; ostia $5.1815 / 7.5442\text{ mm}$), with $2.7696 \pm 1.1919\text{ mm}$ pooled $N=64$ $\text{CD}_{\text{CL}}$ and $128/128$ trees ($\beta_0 = 1, \beta_1 = 0$, $0.0000\text{ mm}$ detachment).
Abstract (translated)
URL
https://arxiv.org/abs/2610.05900
PDF
-
Whole-Body Humanoid Control via Stochastic Motion Tokenization
2026-10-27 00:58:02Michael Drolet, Oleg Arenz, Jan PetersAbstract
While velocity-commanded RL policies are effective for controlling quadrupeds, they are a limited interface for humanoids that need to perform complex whole-body motions. Furthermore, RL tracking policies for whole-body motions typically have limited information about the desired future motion, preventing them from performing controlled deviations to achieve long-term task success; such information is particularly relevant when a controller must prepare for transitions between behaviors. We address these challenges using concise motion representations that can be generated by a high-level controller, given a behavior label or text command. These representations can be fed directly into a tracking policy to provide a detailed motion reference. Our method exhibits a high degree of flexibility in performing real-world robot motions such as dancing, fighting, or walking based on conditional information. At matched capacity, our adopted stochastic motion tokenizer reconstructs motions from the Unitree G1 at a level that exceeds state-of-the-art approaches such as finite scalar quantization (FSQ) and the vector-quantized autoencoder (VQ-VAE). The resulting discrete interface conditions whole-body tracking policies that execute generated dance, fight, and walk motions lasting 16.8 s each and a 35 s fight-to-dance sequence on a physical Unitree G1.
Abstract (translated)
URL
https://arxiv.org/abs/2610.36452
PDF
-
CrowdCC: Jointly Benchmarking Corruption Robustness and Deployment Quantization in Crowd Counting
Abstract
Crowd-counting models deployed on edge devices face two shifts: inputs are degraded along the capture and transmission pipeline, while models are quantized to satisfy resource constraints. Yet corruption robustness and deployment quantization are typically evaluated separately. We introduce CrowdCC, a controlled corruption benchmark that jointly evaluates these effects using 23 deployment-motivated corruptions at three severity levels, four crowd-counting datasets, and 26 models using A100 TensorRT engines at FP32, FP16, and post-training INT8 precision, paired with throughput and energy measurements from corresponding Jetson Orin NX configurations. The corruptions span six families, with head annotations adjusted when geometry or visibility changes. Our evaluation shows that capture-chain corruptions consistently produce the largest degradation across the evaluated models, while crowd density affects absolute and relative error differently. Clean accuracy does not fully predict corrupted performance. INT8 alters corruption robustness substantially more than FP16, with no systematic direction, and does not consistently improve latency or energy. INT8 inference efficiency is strongly associated with reformat overhead in the compiled engine, while slowdowns are also observed under limited INT8 coverage. Precision changes corrupted-error rankings and robustness-cost frontiers, showing why model selection must consider corruption robustness and measured deployment cost together. The code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.36340
PDF
-
Training-Free Inference Acceleration for Video Large Language Models: Methods, Trade-offs, and Evaluation
2026-10-26 21:55:44Yuqi Li, Chuanyue Yan, Neil Rodrigues, Xianglong Wang, Xiaoqin Feng, Junming Liu, Xi Xiao, Zhongwei Cheng, Wei Li, Lin Chen, Yingli TianAbstract
Video Large Language Models (VLMs) have demonstrated strong capabilities in video understanding and reasoning, but processing long visual sequences incurs substantial computational, latency, and memory overhead. This survey examines training-free inference acceleration, focusing on reducing inference costs while preserving task-relevant spatiotemporal evidence. We organize existing methods by their intervention stages in the inference pipeline, covering input, visual encoding, projector, prefill, and decoding. Post-training quantization (PTQ) is discussed separately as a cross-stage extension with varying calibration requirements. We analyze how intervention position affects evidence preservation and computational savings, and why local efficiency improvements may not translate into end-to-end acceleration. We complement this analysis with quantitative evidence from prior studies and our own experiments under a shared evaluation setting to assess accuracy-latency trade-offs.
Abstract (translated)
URL
https://arxiv.org/abs/2610.36327
PDF
-
Quantitative Generalization Bounds for Multi-Layer Transformers: A Few Long Strings can Beat Many Short Ones
2026-10-26 16:36:27Marius Henry, Michael Rizvi-Martel, Michael HahnAbstract
Transformer-based language models exhibit an astonishing capacity to generalize to longer and more difficult problem instances than those seen in training. However, theoretical understanding of this phenomenon still lags behind. Unlike most machine learning models, the generalisation abilities of transformers inherently require considering two axes. On one hand, one must examine in-distribution statistical generalisation, which concerns the ability of a model to broaden a function from training samples to novel samples of the same distribution. On the other hand, given the sequential nature of language data, one must also consider length generalisation, which reflects whether a model trained on inputs of a given length can extrapolate to longer this http URL two notions have mostly been studied separately, leaving open how they interact. In this work, we give a joint treatment of both axes for fixed-precision transformers. We give generalisation guarantees in terms of the parameter sizes of the involved models, as well as the quantization space and the alphabet size that supply the sequences. Finally, we illustrate how the requirements in the sample count and training length trade off. In particular, we find that increasing training length can be exponentially more efficient than increasing the sample size. Experiments support predictions of the theory.
Abstract (translated)
URL
https://arxiv.org/abs/2610.35715
PDF
-
Let the Gate Decide: Per-Lookup Precision for Quantized Conditional Memory in LLMs
2026-10-26 11:53:58Tianqi Chen, Xunchao Li, Linchuan Xie, Guanghua Yu, Jianchen Zhu, Peisong Wang, Jian ChengAbstract
Conditional memory layers, such as DeepSeek's Engram, attach large hashed n-gram tables to a language model, and removing them from DeepSeek-V4.1-Flash costs 36.8 PopQA points. Low-bit quantization loses long-tail facts first, and the standard remedy, giving more bits to frequently accessed rows, backfires: refining lookups by access frequency is worse than refining them at random. We trace this to the gate that scales each retrieved row. A lookup's quantization error is weighted by its gate, which is small for frequent n-grams, large for the rare n-grams of long-tail entities, and computable from a 1-bit code. GateRQ therefore stores each row as a ladder of 1-bit residual codes, keeps level 1 in GPU memory to compute the gate, and reads deeper levels only for the lookups the gate ranks highest. Without retraining, GateRQ cuts GPU table memory by 7.8x on DeepSeek-V4.1-Flash, which raises vLLM throughput by 11%-31% over FP8 tables, while losing about 1 PopQA point at 1.50 bits read per lookup and 0.2 points at 2.10 bits across two models.
Abstract (translated)
URL
https://arxiv.org/abs/2610.35985
PDF
-
AIPC: Evidence-Guided Agent Workflows for PyTorch Project Deployment
2026-10-26 06:10:01Zhanwei Wu (Qualcomm Technologies, Inc), Jianhao Su (Qualcomm Technologies, Inc), Brian Zhang (Qualcomm Technologies, Inc), ShengTing Huang (Qualcomm Technologies, Inc), Weidong Feng (Qualcomm Technologies, Inc), Gang Liu (Qualcomm Technologies, Inc), Zhiguo Li (Qualcomm Technologies, Inc), Wei Huang (Qualcomm Technologies, Inc), Li Tan (Qualcomm Technologies, Inc), Wanjun Li (Qualcomm Technologies, Inc), Zejiong Teng (Qualcomm Technologies, Inc), Changgeng Zhong (Qualcomm Technologies, Inc), Guanping Wu (Qualcomm Technologies, Inc), Han-Hsing Tu (Independent Researcher)Abstract
Deploying and optimizing a PyTorch inference project on a dedicated accelerator, especially an on-device NPU, requires more than using a framework device API as with CUDA. The project must be converted into hardware-runtime artifacts and adapted to the target toolchain and application interface. Differences in model formats, quantization flows, operator support, and toolchain interfaces create cross-layer problems spanning PyTorch/ONNX models, toolchains, runtimes, and device execution. LLM Agents can invoke tools to address such problems, but may drift because of insufficient domain knowledge, large configuration and repair spaces, or poor path selection. We propose AIPC (AI Porting & Conversion), an Agent method for deploying PyTorch inference projects. AIPC combines structured Skills with toolchain, runtime, and profiling evidence to drive cross-layer diagnosis and repair, organizing model adaptation, artifact building, on-target execution, accuracy repair, and performance optimization into one deployment flow. We evaluate AIPC on the on-device NPU path of Qualcomm QAIRT across vision, speech, LLM, and diffusion models. In two optimization cases, 6/7 Skill-arm Whisper-Small runs meet predefined accuracy and latency criteria versus 0/6 control runs, while YOLO-Worldv2 achieves a 2.791x end-to-end speedup versus 1.599x for the control. Under a pre-fixed deployment contract, Skill, tool permissions, and check conditions, some runs achieve zero in-run human intervention (0-HITL). We also examine Qwen3.8-27B as a locally deployable model for driving an Agent workflow.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34890
PDF
-
ThinkBit: Search Codewords, Recalibrate Scales, Select by Reasoning
2026-10-26 05:56:35Rhea Sanjay Sukthanker, Pashmina CameronAbstract
Reasoning language models generate thousands of tokens per answer, and generation speed is limited by memory bandwidth, so storing weights in two bits instead of bfloat16 can potentially make inference 8$\times$ cheaper. Two-bit methods are usually judged by perplexity and multiple-choice question-answering accuracy, both of which say little about whether a model can still reason through more difficult problems. We present ThinkBit, a strictly scalar two-bit framework built to preserve reasoning, in which every weight is an integer codeword times a bfloat16 scale shared by a group of weights. We first use a sensitivity metric to assign each layer a ternary, 2-bit, or 3-bit tier to allocate bitwidths optimally among layers. This is followed by a layerwise Gumbel-softmax-based scalar quantization algorithm that searches for optimal codewords block by block. Since a layerwise scheme does not capture how quantized layers interact with each other, we recalibrate scales, with the codewords frozen, using a knowledge-distillation-based loss against the teacher model. We then perform a short quantization-aware training (QAT) stage that lets the model mimic the base model's distribution by reassigning codewords. Finally, a paired reasoning test on training problems selects the final checkpoint between the recalibration stage and the QAT stage. On Qwen3-1.7B at matched size, our recalibrated checkpoint scores 46.7 on MATH-500 against 30.0 for EfficientQAT trained on the same data, and 73.0 after QAT on reasoning traces; at 0.52 GB it is smaller than every published 2-bit GGUF of the model yet more accurate on MATH-500. We show that this framework scales and up to 14B, a 3.74 GB model scores 94.8 on MATH-500, against 97.5 for the 29.5 GB bfloat16 model.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34886
PDF
-
From Sentinel-2 to 1-m Imagery: Real-World $times 10$ Cross-Sensor Super-Resolution via Visual Autoregressive Modeling
2026-10-26 02:43:59Tailai Chen, Yuluxin Fu, Jie Deng, Shu Luo, Yanan Wu, Danfeng HongAbstract
Reconstructing 1-m imagery from Sentinel-2 combines two difficulties. The input and the reference are observed by different satellites, which differ in spatial resolution, spectral response, radiometry, acquisition time, and local geometry, so no synthetic degradation operator links them. The required magnification of ten in each spatial dimension also far exceeds the factors of two to four at which remote sensing super-resolution is usually developed. We study this real-world $\times 10$ problem using YREB-SR, which pairs 2022 Sentinel-2 RGB composites with 2022 Jilin-1 RGB references resampled to a 1-m grid over the Yangtze River Economic Belt. The benchmark contains 11,559 geographically matched scene pairs, with 6,409, 2,135, and 3,015 assigned to training, validation, and testing, respectively. We develop CrossVAR, a visual autoregressive model that predicts the 1-m latent as a nine-scale hierarchy conditioned on the Sentinel-2 input and estimates the remaining quantization residual with a diffusion refiner run without injected noise. Complete 1600 $\times$ 1600 scenes are reconstructed in overlapping windows that start on the Sentinel-2 pixel grid, as every training patch does, and every window is decoded at the size on which the decoder was trained before the windows are blended with Gaussian weights. On all 3,015 test scenes, CrossVAR achieves 27.60 dB PSNR, 0.7725 SSIM, 0.1451 LPIPS, and 7.15 FID, the best value of each metric among the compared methods. These results establish a strong autoregressive baseline for large-scale cross-sensor reconstruction. To facilitate reproducibility, the source code and implementation are publicly available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34742
PDF
-
Real-Time Conveyor Belt Defect Detection via Edge TPU-Accelerated YOLOv11 on Raspberry Pi 5
2026-10-25 14:14:37Muhammad Nur Bijak Bestari, Mulki Indana ZulfaAbstract
Conveyor belts are vital arteries in continuous bulk material handling, particularly in heavy process industries such as cement manufacturing. Undetected structural anomalies (longitudinal tears, holes, degraded patch repairs, and impact punctures) cause catastrophic belt snapping, costly downtime, and severe safety hazards. While deep learning offers high accuracy, existing models rely on power-hungry workstation GPUs, hindering practical deployment at remote industrial edge points. In this paper, we propose, implement, and validate an end-to-end, low-power edge AI vision system for real-time conveyor belt defect detection deployed on a Raspberry Pi 5 accelerated by a Google Coral Edge TPU coprocessor. The system leverages an optimized YOLOv11 nano (YOLOv11n) architecture trained on a multi-source industrial corpus exceeding 3,000 augmented samples from an operational cement plant in West Java, open-source datasets, and automated captures. The model undergoes post-training full integer quantization (INT8) compiled for systolic tensor execution. Experimental results demonstrate robust detection capability, achieving precision of 0.742, recall of 0.730, and mAP@0.5 of 0.764 across four defect classes. Deployed on-site in an IP-rated enclosure interfaced with an RTSP PoE industrial camera, the system achieves stable throughput of 30-40 FPS at 320x320 resolution with an end-to-end latency of 70-90 ms. Operational profiling under harsh clinker dust confirms thermal equilibrium at 55 +/- 5 deg C on the host CPU and ~65 deg C on the Edge TPU with zero frame loss. This demonstrates that edge AI bridges the gap between high-accuracy computer vision and cost-effective predictive maintenance in continuous manufacturing.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34304
PDF
-
ResCom: Activation Residual Compensation for Large Language Model Quantization
2026-10-25 03:02:21Ning Yang, Bohan Cui, Zhuowen Liu, Yuting WangAbstract
Low-bit weight--activation quantization reduces the memory and computation needed to deploy large language models, but activation rounding can still degrade accuracy after calibration. Reconstruction from rounded inputs cannot distinguish activations within the same quantization bin. We propose ResCom, a low-rank correction that uses the activation rounding residual to compensate for the resulting layer-output error. An output-aware calibration procedure selects residual directions by their contribution to output reconstruction and computes the correction in closed form, without gradient training or updates to the calibrated weights. On Llama-3-8B with 4-bit weights and activations, rank-32 attention corrections reduce WikiText-2 perplexity from 7.708 to 7.147, improve MMLU by 2.89 percentage points, and raise five-shot GSM8K accuracy from 23.96 to 36.16 percent. Evaluations on models from 1B to 70B parameters and on held-out C4 test the method beyond this setting. The evaluated FP32 implementation adds 2.25 percent to 4K prefill time; a separate compact-storage implementation retains a 27.7 percent peak-memory reduction relative to FP16.
Abstract (translated)
URL
https://arxiv.org/abs/2610.35955
PDF
-
QECGBERT: A Scalable ECG Foundation Model Based on Self-Supervised Language Learning via VQ-VAE Waveform Tokenization
2026-10-25 00:37:53Aruna Mohan, Burcu Ozek, Or Zilbershot, Fatemeh AfghahAbstract
AI foundation models play a pivotal role in advancing heart disease diagnosis from electrocardiogram (ECG) data, particularly for rare diseases where the lack of sufficient labeled data hinders supervised learning. However, most self-supervised learning methods treat ECGs as generic sequence or image data without exploiting underlying physiological structure. To address this limitation, we propose QECGBERT, a self-supervised ECG language model that explicitly encodes physiological semantics by tokenizing the P, QRS, T, and isoelectric background waveforms. Vector-quantized variational autoencoder (VQ-VAE) tokenization is applied to each 12-lead ECG, with tokens enriched by continuous waveform features to preserve morphology lost during quantization without increasing sequence length. This design enables compact BERT masked language modeling with short sequences, improving scalability and reducing computational cost relative to long-sequence approaches. QECGBERT applies BERT pre-training to the MIMIC-IV dataset containing over 800,000 unlabeled ECG recordings, and fine-tuning on labeled PTB-XL, CPSC2018, and Chapman--Shaoxing--Ningbo datasets for evaluation. The results demonstrate that QECGBERT achieves comparable or superior performance to published benchmarks, offering a scalable, physiologically aware approach to ECG foundation modeling for cardiac diagnosis.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33964
PDF
-
Adapting DNNs to Embrace Voltage Scaling Induced Errors via Accumulation-Aware Quantization
2026-10-24 17:42:49Alexis Maras, Panagiotis Chaidos, Georgios Alexandris, Dimitrios Soudris, Sotirios XydisAbstract
The pervasive deployment of Deep Neural Networks (DNNs) across the computing spectrum has brought growing attention to energy efficiency, alongside raw performance. While voltage scaling is a potent knob for power reduction, its application in next-generation nodes is hampered by severe throughput degradation and susceptibility to process variation. This work aims to address these energy-reliability trade-offs by introducing a hardware-software co-design methodology, based on the observation that timing-robust bit ranges within the accumulator of a systolic array vary significantly across different voltage-frequency operating points. At the software level, we propose an Accumulation-Aware Training approach that incorporates this information as a first-order optimization constraint during training. By integrating mixed-precision quantization with unstructured pruning, we effectively constrain the model's weight distribution to avoid activating the hardware's sensitized high-rank carry chains. At the architectural level, we complement this with selective MSB-masking and fine-grained clock-gating to minimize unnecessary dynamic switching activity. Furthermore, we employ a layer-wise Dynamic Frequency Scaling strategy to further adapt to the accumulator's reduced dynamic range, based on the numerical sensitivity of each layer. We experimentally evaluate the proposed solution on a deep scaled 2nm technology node and demonstrate that, when operating at a low-voltage frontier, our co-optimized post-layout architecture delivers substantial energy and energy-delay product gains of 39% and 23% respectively, relative to the nominal voltage implementation, while maintaining accuracy degradation within 1%.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33824
PDF
-
RAQ-MLA: Role-Aware Quantization for Multi-Head Latent Attention Caches
2026-10-24 11:54:55Tianqi Chen, Xunchao Li, Linchuan Xie, Guanghua Yu, Jianchen Zhu, Peisong Wang, Jian ChengAbstract
Multi-Head Latent Attention (MLA) has been adopted by recent frontier large language models. Instead of per-head keys and values, MLA caches a compressed KV latent that is up-projected into keys and values at decode time, together with a narrow positional key that carries position information into the attention score. Existing KV-cache quantization methods are designed around the keys and values of standard attention and do not account for what MLA actually caches, which costs accuracy at low bit widths. We analyze the MLA cache both in activation distribution and in how quantization error propagates to the attention output, and show that the two cached tensors play different roles: the latent mainly supplies the values, whereas the positional key dominates the attention score. We therefore propose RAQ-MLA (Role-Aware Quantization for Multi-Head Latent Attention Caches), which fits an invertible transform to each cached tensor against the error that tensor delivers to the attention output: attention-weighted value error for the latent and score perturbation for the positional key. Without training, RAQ-MLA quantizes the full MLA cache to about 2.5 bits per element, a 6.5x increase in theoretical KV-cache capacity, while remaining nearly lossless on reasoning and coding tasks across four MLA models.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33625
PDF
-
MOSHO: A Native Discrete Shark-Inspired Metaheuristic for Hardware-Aware Neural Architecture Search
2026-10-23 22:56:01Nazim Abdeerhmane Aouni, Zakaria Soualah Mohammed, Maroua Ogab, Nour El Imane Elbar, Rayane Rahmat Errahmane Smara, Billel Moussous, Adriane Anis Khaled, Malika Bessedik, Fatima Benbouzid Si Tayeb, Hadjer BenmezianeAbstract
Hardware-aware neural architecture search (NAS) requires jointly optimizing predictive accuracy and device-specific latency over discrete search spaces. While multi-objective swarm optimizers show promise, continuous formulations such as the Multi-Objective White Shark Optimizer (MOWSO) can suffer from quantization noise when rounded to integer-encoded architecture vectors. To address this limitation, we propose MOSHO (Multi-Objective Shark Hunting Optimization), a native discrete metaheuristic that replaces continuous motion equations with discrete architecture-editing operators. MOSHO integrates adaptive operator scheduling, energy-regulated restarts, and deferred crowding-distance truncation to enhance exploration and search efficiency. On the merged NAS-Bench-201/HW-NAS-Bench benchmark across 18 contexts, MOSHO achieves a median normalized hypervolume of 0.9987 and IGD+ of 0.0064, yielding Pareto fronts competitive with NSGA-II while reducing search runtime by 96.4% (0.64 s vs. 17.72 s). These results demonstrate that tailored discrete search mechanisms can enable swarm optimizers to achieve high-quality multi-objective solution quality with substantially lower search cost in hardware-aware NAS.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33358
PDF
-
TinyFlow: Breaking the Warping Bottleneck for Edge Optical Flow
2026-10-23 22:21:29Xiao'ao Song, Chahat Deep SinghAbstract
High-performance optical flow networks are difficult to deploy on resource-constrained edge accelerators because their iterative refinement warps features by the current flow, which relies on flow-dependent resampling and data-dependent memory access. We present TinyFlow, a 1.26M-parameter warp-free optical flow network that preserves iterative coarse-to-fine refinement by replacing the warp with a fixed-window correlation over constant offsets, so the whole network compiles as a single static INT8 graph. Quantization-aware fine-tuning, warp-teacher distillation, and model souping reduce the Float32-to-INT8 penalty to below 0.2 end-point error (EPE) on MPI-Sintel. TinyFlow outperforms EdgeFlowNet and NanoFlowNet on Sintel Clean, Sintel Final, and FlyingChairs in both Float32 and INT8, while using 54% fewer parameters than EdgeFlowNet, and our INT8 model achieves 6.29/7.13 EPE on Sintel Clean/Final. In flow-only monocular visual odometry on TartanAir and UZH-FPV, the same INT8 model yields the lowest trajectory error on seven of eleven sequences.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33338
PDF
-
From Perception to Pixels: Human-Derived Visually Distinct Colors for Improved Dominant Color Extraction
2026-10-23 10:42:23Uldana Shyndali, Nurdaulet Toleugaliyev, Kuat Zharykbayev, Zarina Abildasheva, Pakizar ShamoiAbstract
Dominant color extraction is a fundamental task in computer vision, with applications in design, image retrieval, and content recommendation. However, conventional algorithms based on device-oriented color spaces (e.g., RGB) and fixed quantization often fail to capture how humans perceive color differences. In this study, we propose a human-perception-based framework for dominant color extraction, built on a novel Human Color Dictionary (HCD) derived from a perceptual experiment. Participants (n=110) were sequentially shown 120 colors and asked to identify those they perceived as unique. Based on their responses, we constructed a 120 by 120 co-occurrence matrix and applied a greedy selection algorithm to derive 64 perceptually distinct colors, forming the HCD. This dictionary was integrated into RGB color histogram and K-means color extraction algorithms by initializing centroids and mapping pixel values to their nearest perceptual color using the deltaE 00 distance. Evaluation on a dataset of 4,562 images paired with expert-designed palettes demonstrated that our method improves perceptual alignment, as measured by Earth Mover's Distance (EMD) and best-match distance. Our findings suggest no meaningful gender differences in the number of uniquely perceived colors, indicating that sex does not significantly influence color perception in this task. Another key finding is that highly saturated colors are perceived as the most distinguishable, whereas dark and deep colors tend to perceptually collapse and are less reliably differentiated. The results confirm that incorporating experimentally derived human color representations ensures the generated color clusters align more closely with human perception, thus improving the usability of color palette generation in practical applications.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32473
PDF
