arXiv subjects

Ang Li

Publications and source records attributed to Ang Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

HeteroVDiff: Pipelining Video Diffusion across Mobile NPU and GPU with Shared Weights

Video generation is moving from the cloud to phones for privacy and offline use, and applications often ask for several clips at a time, so on-device video generation is a throughput problem before it is a latency problem. Contemporary mobile SoCs carry both a GPU and a neural processing unit (NPU) on shared DRAM, yet existing on-device video diffusion systems run the denoiser on one engine while the other sits idle or only decodes the finished latent afterwards. Heterogeneous designs built for mobile LLMs do not carry over: they synchronise the two engines at every operator and partition weight tensors between them instead of sharing one stored copy. We first measure the platform limits that decide what can run on this class of SoC: how large one buffer may be, whether both runtimes can map the same buffer, and how much activation memory the NPU toolchain lets one graph hold. Guided by these measurements, we store the base weights once in INT8 in the layout the NPU reads natively; the GPU consumes the same bytes directly as integer dot-product operands, so neither runtime rearranges weights on load. We then flatten the denoising trajectory into its full sequence of block executions and cut it once at a block boundary, so the GPU runs the prefix and the NPU the suffix. The cut is placed from measured rather than modelled rates, including the decoder that only the NPU can run, and taken while both engines run together because each engine's speed changes when the other is active. We present HeteroVDiff, which serves a multi-expert video diffusion model across the NPU and GPU with 48% less weight memory than giving each engine its own copy. Generating videos back to back on a Galaxy S25 with decoding inside the pipeline, HeteroVDiff sustains 1.35x the stream throughput of NPU-only execution, and on VBench-2.0 it matches FP32 in the total score over all 18 dimensions.

cs.DC↗

arXiv2026

CivWorldBench: Evaluating Language World Models in Structured Multi-Agent Sandbox Worlds

The growing use of language models for forecasting motivates a broader evaluation of their world-model capabilities. Such evaluation must go beyond predictive accuracy to assess hidden-state inference and intervention prediction. Real-world forecasting benchmarks offer limited access to hidden-state ground truth and controlled experiments. We introduce CivWorldBench, a benchmark built on a multi-agent strategy game engine with recorded internal states, controlled observations, and paired intervention rollouts. Its 1,900 decision points from 475 worlds support three task levels: association forecasting, belief inference, and intervention prediction. Reference predictors support attributable evaluation of capability gaps, with an expert harness and empirical ceiling estimates contextualizing scores. Stored trajectories make intermediate predictions verifiable, while agent-play experiments test whether world-model capabilities support decision-making. The main panel comprises eleven frontier language models. Composite benchmark scores correlate with real-world forecasting leaderboards on matched model panels, more closely than with a general-capability index, supporting its use as a proxy for real-world forecasting capability. Strong association forecasts do not consistently extend to belief inference or intervention prediction, and effect magnitudes remain poorly calibrated despite informative direction predictions. Process verification reveals errors missed by final-answer evaluation, and structured reasoning improves intervention prediction for the strongest model in the process study. Agent-play experiments identify useful assisted plans and motivate evaluating decision-relevant capabilities beyond association forecasting. These findings identify hidden-state inference, intermediate-state prediction, and effect calibration as priorities for world-model training.

cs.CL↗

arXiv2026

CacheFlex: Direct Software-Managed Access to Higher-Level Cache for Scalable Vector Support

Wider vector/SIMD lanes multiply compute throughput, but cache improvements remain challenging. We identify L1 data cache capacity as the most impactful parameter for performance and efficiency. L1 capacity is constrained by pipeline coupling and latency requirements; private L2 offers an order of magnitude more capacity but is inaccessible from software. This work proposes CacheFlex, an ISA and microarchitecture co-design that enables software-managed, direct access to private L2 SRAM by allowing runtime conversion of parts of L2 into a scratchpad memory (SPM). Through a lightweight ISA extension, software explicitly controls data placement and lifetime within the L2. CacheFlex integrates into an out-of-order pipeline with cross-domain memory consistency and OS context-switch support, with merely 2.02% area overhead relative to the L2 in a post-layout 12 nm implementation. We prototype CacheFlex based on the Armv8-A SVE ISA, model it in gem5, and evaluate it using optimized general matrix multiplication (GEMM) kernels with diverse model-derived shapes, FlashAttention-based fused attention, and full-layer workloads from LLaMA-3.2-1B and BERT-Base in the prefill phase. As the vector length (VL) scales from VL = 4 (512-bit) to VL = 16 (2048-bit), CacheFlex's kernel-level geometric-mean speedup and memory-subsystem energy reduction grow from 1.12x and 26% to 1.44x and 40%, respectively, bridging the widening gap between per-instruction data demand and memory performance. On full transformer layers combining CacheFlex with FlashAttention, the design delivers up to 1.78x speedup and 45% memory-subsystem energy savings (BERT-Base, sequence length T=4096, VL = 16; LLaMA-3.2-1B: 1.64x and 38%). CacheFlex is complementary to hardware prefetching, and its benefit grows as vector width and peak compute throughput increase for compute-intensive AI kernels.

cs.AR↗

arXiv2026

GPU-Based Parallel BP Decoding for Quantum LDPC Memory

Quantum low-density parity-check (qLDPC) decoding requires reliable corrections and low latency over long syndrome histories. We extend BP-SF, an oscillation-guided belief-propagation (BP) decoder with independent syndrome-flip trials, by introducing windowed and streaming operations. Our CUDA implementation combines disordered-memory retries, sliding windows, correlation-aware graphs, and layered scheduling. Circuit-level simulations test fixed-basis memory and complete logical Pauli-frame recovery. On the [[144,12,12]] bivariate bicycle code at physical error rate p = 0.002, parallel BP-SF reduces 99.9th-percentile call latency by a factor of 13.15 relative to an eight-chain Relay-BP GPU baseline on 100,000 matched samples for which every tested method's initial BP stage fails within 80 iterations. Windowed BP-SF maintains a flat logical error rate and nearly constant per-round compute time through 512 rounds, reaching a 17.1-fold speedup over monolithic decoding. In 512-round full-frame streaming decoding with 2.97-ms syndrome arrivals, layered BP-SF on a transformed joint graph with 20 retries reduces 99.9th-percentile correction-release age from 166.29 to 36.25 ms relative to the joint graph with flooding updates and 100 retries. These results connect parallel BP post-processing to lower decoding costs and earlier correction release in streaming quantum memory.

quant-ph↗

arXiv2026

Model-independent ZH production cross section at FCC-ee

This paper presents prospects for measuring the model-independent ZH production cross section at the FCC-ee using the recoil-mass method at center-of-mass energies of 240 GeV and 365 GeV. Analyses are carried out in the muon, electron, and hadronic decay modes of the associated Z boson. The event selections rely primarily on the kinematics of the reconstructed Z decay products, ensuring maximal independence from specific Higgs boson decay modes, while multivariate techniques are employed to further enhance sensitivity. The statistical interpretation of the leptonic and hadronic final states at 240 GeV, with an integrated luminosity of 10.8 ab$^{-1}$, yields relative precisions of 0.52% for the combined leptonic channels and 0.38% for the hadronic channel. Their full statistical combination leads to uncertainties of 0.31% at 240 GeV and 0.52% at 365 GeV, the latter assuming an integrated luminosity of 3.12 ab$^{-1}$. Dedicated statistical tests demonstrate model independence at the level of the obtained precision. This study presents the first consistent and combined analysis of the leptonic and hadronic final states for a model-independent ZH cross-section measurement at a future lepton collider, using a unified workflow and covering both $\sqrt{s} = 240$ and 365 GeV. It provides the most precise expected measurement of the ZH production cross section at future lepton colliders, with the degree of model independence demonstrated within the achieved precision.

hep-ex↗

arXiv2026

PAIR: Bridging Perception and Action in Vision-Language-Action Models

Vision-language-action (VLA) models map visual observations and language instructions to continuous robot actions. This task requires a transition from representations that describe the scene and instruction to representations that support action generation. Many continuous-action VLAs leave this transition implicit and supervise it mainly through the final action-prediction loss. We introduce PAIR, a framework that learns a shared perception-action representation between these two spaces. During training, a Masked Action Autoencoder encodes expert action chunks into horizon-aligned Action Latent Tokens. A Bridge Module extracts task-relevant features from the current visual-language representations. PAIR aligns these features with the Action Latent Tokens to form Bridge Tokens that preserve task information and capture the structure of expert actions. The Bridge Tokens are then projected into the action-token space and injected into the initial Action Tokens, providing an action-ready starting point for Action Expert refinement. At inference, the autoencoder is removed, and the Bridge Tokens are generated only from the current observation and instruction. Experiments on LIBERO, LIBERO-Plus, and CALVIN ABC-D show gains for the evaluated OpenVLA-OFT and VLA-Adapter models. On LIBERO-Plus, PAIR raises VLA-Adapter's success rate from 59.1% to 64.2%. On CALVIN, it increases VLA-Adapter's average completed sequence length from 4.42 to 4.53. Across seven real-world tasks, PAIR raises OpenVLA-OFT's success rate from 51.4% to 65.0%. Representation analyses show that Bridge Tokens retain task information while making continuous-action information accessible before Action Expert refinement. These results support a shared intermediate representation as a useful interface between perception and action in continuous-action VLAs.

cs.AI↗

arXiv2026

DIVA: Dual-Space Intent-Aware Visual Attenuation for Vision-Language-Action Policies

Vision-language-action (VLA) policies typically feed dense visual patch tokens into a language-action backbone, preserving scene context but offering no explicit mechanism to regulate how strongly different visual tokens influence policy computation. We introduce DIVA, a Dual-Space Intent-Aware Visual Attenuation module with an anchor-then-attenuate design. DIVA combines high-level task intent with low-level visual evidence to estimate patch-wise relevance anchors, then applies them in two complementary spaces: it reweights projected visual tokens before backbone entry and persistently attenuates low-relevance visual states within the backbone. DIVA preserves the full visual token sequence and requires no external grounding supervision. On LIBERO, DIVA improves OpenVLA-OFT from 96.6% to 98.0% average success and raises its zero-shot LIBERO-Plus score from 69.6 to 72.6. Real-world experiments further show consistent gains under task-irrelevant visual perturbations, supporting the robustness of intent-aware visual attenuation beyond simulation.

cs.AI↗

arXiv2026

Interval Deformations of Coxeter Arrangements

We study integer-interval deformations of Coxeter arrangements of types $A$, $B$, and $D$, assigning an integer interval to each root direction through an integer-valued root function. We introduce a local compatibility condition on this root function. For compatible root functions, subject to additional local bounds in types $B$ and $D$, we derive characteristic-polynomial formulas as binomial sums over cyclic orders in type $A$ and signed permutations in types $B$ and $D$. The proofs use the finite-field method and reduce hyperplane avoidance to consecutive-gap inequalities. These formulas recover, in particular, the classical Catalan and Shi characteristic polynomials.

math.CO↗

arXiv2026

Improved Convergence of Large Stepsize Gradient Descent for Logistic Regression

We study gradient descent (GD) with a large constant stepsize for logistic regression on linearly separable data. Existing analysis shows an accelerated rate of $\widetilde{O}(1/\sqrtε)$ to reach loss $ε$ with an aggressive stepsize, although the loss may initially oscillate. Tighter control of the oscillatory dynamics has been available only for two-dimensional data. We prove a substantially faster rate in arbitrary dimension: GD with a large stepsize $η=1/ε$ reaches loss $ε$ within $O(\ln^{p}(1/ε))$ steps, where $p$ depends only on the margin and the rank of the data. Our proof improves the bound on the transition time of GD from the oscillatory to the stable phase, after which the loss decreases monotonically. We split the oscillatory phase into recursively nested intervals. The margin and the rank bound the nesting depth, and a counting argument bounds the number of intervals at each depth, together yielding the polylogarithmic step complexity.

cs.LG↗

arXiv2026

Ask, Relax, or Act? Evaluating Actionable Indeterminacy in LLM Preference Reasoning

An LLM agent can recognize uncertainty yet still choose the wrong next step: asking when action is already justified, or seeking clarification when the constraints must change. We formalize actionable indeterminacy: act when an accepted action is shared across all admissible preferences or objectives, clarify when each possibility is feasible but no action is shared, and propose a minimum-cost permitted constraint repair when the request is infeasible. We construct a solver-grounded benchmark spanning object allocation, meeting scheduling, apartment choice, and stable matching. Matched pairs retain the same source while changing whether intervention is necessary, and evaluation separates decision correctness, matched-pair reliability, and fully correct responses. Our findings reveal a recurring difficulty in recognizing when intervention is unnecessary: models can identify situations requiring clarification or repair yet still intervene when a justified action already exists. Correct decision labels also fail to guarantee usable actions, questions, or repairs. Crucially, response requirements shape not only how decisions are expressed but also which decisions are made. Making the required content explicit substantially improves fully correct responses and can change intervention decisions, even when outputs are already parseable. These findings highlight that reliable agency requires more than recognizing uncertainty: it requires intervening only when necessary and translating the chosen next step into a verifiable response.

cs.CL↗

arXiv2026

XGenAct: Geometry-Enhanced World Action Models through Cross-Task Generation

World action models (WAMs) have advanced robot control by predicting how observations and actions evolve over time. Despite this progress, RGB and action based future prediction does not explicitly address the spatial understanding needed for robot manipulation. Existing efforts often add a limited set of spatial prediction tasks through specialized heads or branches, leaving both the range of spatial supervision and the model architecture fragmented. We introduce XGenAct, a world action model that represents RGB observations, robot actions, metric depth, surface normals, and functional role segmentation as RGB videos through deterministic codecs. By sampling perception and action tasks during training, XGenAct uses one video diffusion transformer and one objective to learn temporal prediction across these spaces without modality specific learned heads. On held out RLBench tasks, structured perception training improves average closed loop success over RGB only training, and XGenAct achieves 52% success in the five task external comparison, versus 26% for the strongest evaluated baselines. It also predicts future depth and segmentation more accurately than the evaluated pipelines that generate RGB first and then apply a frozen perception expert.

cs.RO↗

arXiv2026

MoEless: Efficient MoE LLM Serving with Serverless Experts

Large Language Models (LLMs) increasingly adopt Mixture-of-Experts (MoE) architectures to scale efficiently under stringent resource constraints. However, MoE's sparse activation causes severe expert load imbalance, where a few experts become stragglers while others remain underutilized, leading to inflated inference latency and cost. Existing solutions assume static, serverful model deployments, limiting expert elasticity and often incurring costly expert swapping or degraded output quality. We present MoEless, an efficient serverless MoE serving framework that mitigates expert load imbalance via elastic expert execution. MoEless leverages lightweight, layer-aware predictors to estimate incoming expert load distributions and proactively identify stragglers. We design optimized scaling and placement strategies to improve function locality, GPU utilization, and cross-expert load balance. MoEless is prototyped on top of Megatron-LM and deployed on an eight-GPU testbed. Experiments with open-source MoE models and real-world workloads show that MoEless reduces inference latency by 43% and inference cost by 84% compared to state-of-the-art solutions.

cs.DC↗

arXiv2026

Staircase Policy: Streaming Inference for World-Action Models with Large Action Chunks

World-Action Models (WAMs) improve robotic manipulation by conditioning action generation on predicted future observations, but future prediction adds further inference overhead to already expensive iterative action generation. Action chunking can amortize this cost over multiple actions, yet performance degrades over long execution horizons because later actions remain conditioned on stale observations. We introduce STAIRCASE POLICY, a streaming inference and training framework that turns a flow-matching VLA into a JEPA-style WAM and partitions a large action chunk into sub-chunks at staggered denoising stages. Near-term actions are executed as soon as they become available, while later actions continue to be refined. At each sub-chunk boundary, the future latent is re-predicted from the latest observation and used to update all unexecuted actions, enabling long-horizon execution without repeated full policy inference. The resulting future-prediction error can further serve as a signal for adaptive chunking. S-WAM achieves 97.7% on LIBERO and 87.9% on LIBERO-Plus, and improves performance across multiple policy backbones and real-robot tasks. It reaches 292.7 executed actions per second, $3.62\times$ the throughput of conventional execution at comparable accuracy, while reducing time-to-first-action from 123.6 to 73.3 ms. With additional inference optimizations, throughput further increases to 642.9 actions per second.

cs.RO↗

arXiv2026

ARC-KV: Amortizing Anchor Search for Reconstruction-Based KV Cache Compaction

Long-context large language model inference is bottlenecked by KV caches that grow linearly with sequence length. This burden is especially severe for long, reusable context prefixes, whose cache must serve many downstream queries. Reconstruction-based methods such as Attention Matching achieve strong downstream task performance with compact KV caches. However, iterative anchor search dominates the compaction cost of OMP-based Attention Matching. This motivates our selective amortization principle of learning a reusable anchor-selection policy across contexts while retaining context-specific reconstruction. In this work, we propose ARC-KV, a novel reconstruction-based KV cache compaction method that follows this principle. To this end, we first train a value-aware indexer to select real-key anchors in a single scoring pass. ARC-KV then applies convex-hull-constrained key merging and fits an attention-mass bias and compact values against the full cache. At inference time, ARC-KV builds the compact cache once per context using the frozen indexer and reuses it for all subsequent queries. Extensive experiments demonstrate that ARC-KV outperforms reported compaction methods in most settings across QuALITY, RULER, and LongBench on Llama-3.1-8B-Instruct. In particular, at 10% KV retention on QuALITY, ARC-KV improves accuracy from 0.6409 to 0.6474 over Attention Matching while reducing compaction time by a factor of 25.73, from 959.8 s to 37.3 s.

cs.AI↗

arXiv2026

Neuro-Symbolic Computer Use: Learning Reusable Policies for Reliable and Efficient Execution

Many computer tasks recur: the same workflow runs many times, with new inputs and from different starting states. Current computer-use agents re-plan every step of every run, which makes them costly and unreliable on such tasks. We introduce neuro-symbolic computer use, in which a recurring workflow is executed by a learned policy rather than re-derived by an agent on each run. The policy fixes the decisions that are stable across runs (ordering, variables, loops, and branches) in executable code, and delegates observation-dependent decisions, such as grounding and state checks, to neural models. We learn these policies with neuro-symbolic policy iteration: starting from one agent trajectory, it executes the policy, diagnoses failures with task-completion and step-level judges, and revises the code with a coding model informed by an agent's continuation from the point of failure, without access to the benchmark evaluator. Iterating on generated parameter and initial-state variants makes the policy reusable, and a pre-action verifier guards each state-mutating step at deployment. On OSWorld-Verified and ScienceBoard, the learned policies achieve the highest Pass^3 of all methods in all four settings, 3.6-15.8 points above the base agent, while cutting per-run cost by 15-217$\times$ and latency by 3.4-5.1$\times$. On OSWorld-Verified, policies built only on variants transfer to the held-out original tasks, exceeding AutoRPA by 8.6-17.5 points in Pass^3.

cs.AI↗

arXiv2026

SkillCome: Group Contrast Skill Optimization with Dual Memory

Skill evolution improves the capabilities of large language models by analyzing trajectories generated under a given skill and modifying the skill accordingly. Existing approaches typically generate a single trajectory per question. However, this provides insufficient optimization signals since it requires inferring effective skill edits from a solitary path. It is difficult to pinpoint which actions caused the failure in a failed trajectory, or to determine which actions in a successful one should be incorporated into the skill. Furthermore, they rely on a local batch of trajectories for analysis, making the optimization direction susceptible to noisy evidence. To address these, we propose SkillCome, a Skill-evolution method based on group Contrast optimization with dual memory. For each question, SkillCome generates trajectories and performs group contrast analysis to precisely identify key behavioral divergences between successful and failed trajectories, offering reliable optimization signals. The dual memory system further accumulates evidence from historical steps to track patterns shared across different groups, leading to more generalized optimization directions. Together, SkillCome builds a systematic optimization process that transforms experience from observed successful trajectories into reusable skills. Extensive experiments on six benchmarks spanning question answering, reasoning, and agentic tasks demonstrate the effectiveness of our method. SkillCome consistently outperforms baselines across five models of varying families and scales, with gains up to +5.69 points.

cs.AI↗

arXiv2026

Mirage: a Clean-Label Backdoor against LiDAR 3D Object Detection

Deep neural network-based LiDAR 3D object detection serves as a critical perception component in safety-critical autonomous systems. However, recent studies have revealed its vulnerability to backdoor attacks. Existing attacks typically require white-box access or label modification and focus on geometric attacks such as object disappearance or bounding-box manipulation. In this paper, we present Mirage, a black-box and clean-label backdoor attack against deep neural network-based LiDAR 3DOD. Mirage injects a small number of label-consistent poisoning samples into the training set, causing the model to learn a malicious association between a trigger pattern and an attacker-chosen target class while preserving normal training semantics. As a result, the compromised model behaves normally on benign inputs yet systematically misclassifies triggered objects as the target class during deployment. We evaluate Mirage on multiple state-of-the-art LiDAR 3DOD models and benchmark datasets. Experimental results show that Mirage achieves a 73% misclassification success rate with a poisoning rate of only 0.5%, while maintaining detection performance close to that of benign models.

cs.CV↗

arXiv2026

HiMed: Incentivizing Hindi Reasoning in Medical LLMs

Medical large language models hold promise for reducing healthcare disparities, yet Hindi remains severely underrepresented. While medical LLMs excel in high-resource languages, their performance degrades sharply in Hindi, particularly on Indian systems of medicine. We argue that robust cross-lingual medical transfer requires Hindi reasoning. To this end, we introduce HiMed, a Hindi reasoning medical corpus and benchmark suite covering both Western and Indian medicine. We further propose HiMed-8B, a Hindi-form medical reasoning LLM, through the design of decaying scaffolding reward. Extensive experiments demonstrate improvement in Hindi medical reasoning performance and a reduction in the English-Hindi accuracy gap. Ablation studies validate the contribution of each training stage and reward component. All code, data, and weights are available

cs.CL↗