每天知道
你的领域
出了什么

Skills can improve the performance of Large Language Model (LLM) agents by providing task-specific procedural guidance, while skill optimization further improves their effectiveness through iterative refinement. However, existing skill optimization methods typically represent skills as unstructured natural-language instructions, creating two key challenges: 1) Unstructured skills often lack explicit workflow-level guidance and contain substantial redundancy, making them difficult for LLMs to execute; 2) the vast search space of unconstrained natural-language skills makes skill optimization ineffective. To address these challenges, we propose representing skills as graph-structured natural-language artifacts. In graph-structured skills, each node represents an execution step together with its operational guidance, while directed edges encode context-dependent transitions between steps. Compared to unstructured skills, graph-structured skills can provide clear workflow-level guidance. Moreover, the proposed graph-structured skill can also facilitate skill optimization. Building on this structured representation, we introduce GraphSkillEvo, a population-based evolutionary optimization framework with mutation and crossover operators for graph-structured skills. By maintaining multiple candidate skills and combining effective components, GraphSkillEvo enables broader and more comprehensive exploration of the structured skill space than purely LLM-based iterative self-refinement. Extensive experiments across five agent benchmarks demonstrate that GraphSkillEvo consistently outperforms the strong skill optimization baseline SkillOpt, improving average accuracy by 4.01% on GPT-5.4-nano and 1.76% on GPT-5.4. Our code is available at https://github.com/ruisun7/GraphSkillEvo.

Memory systems for large language models have focused predominantly on efficient retrieval, whereas the decision of whether retrieved memories should be trusted has received comparatively little attention. When the memory store contains conflicting positions, standard retrieval-augmented generation (RAG) blindly injects memories and amplifies hallucinations: in models susceptible to memory injection, the RAG hallucination rate under conflicting memories is markedly higher than that of a memory-free baseline. Inspired by memory signaling mechanisms in the prefrontal cortex, we propose the Memory Decision Layer (MDL), a zero-parameter memory decision controller situated between the retrieval and generation stages. Its core is a three-signal complementary encoder that fuses relevance, reliability, and task risk through QR-based orthogonal subspace projection and a meta-working-memory signal into an interpretable decision representation that quantifies the trustworthiness of retrieved memories. Building on this encoder, MDL explicitly decouples confidence from consistency and introduces risk inversion and explicit abstention. Evaluations on mainstream large language models and multiple open-source datasets show that MDL reduces the hallucination rate under conflicting memories by about 56.04% in general scenarios and approaches zero hallucination in high-risk scenarios. The controller is fully white-box: it relies purely on geometric operations, requires no trained parameters, and adds only about 0.14 ms per decision — roughly 50x faster than the embedding-retrieval step that precedes it and four to five orders of magnitude faster than an LLM self-evaluation call.

We present MintAct, a family of vision-language models that unifies UI grounding, multi-step navigation across mobile, desktop, and web, and visual tool use, trained at 2B, 4B, and 8B scales. Through careful design of our environments, data, and training recipes, MintAct models match the performance of per-domain specialists across all of these capabilities. To enable this, we develop a scalable environment and reinforcement learning (RL) infrastructure. On the environment side, we host hundreds of concurrent instances across heterogeneous per-domain backends, serving both trajectory data collection and online RL. To enable efficient and scalable RL training, an asynchronous framework keeps explicit control over the cross-domain training distribution and remains stable under noisy environment feedback and off-policy drift. Experimental results show that MintAct achieves state-of-the-art performance (48.9 on OSWorld-Verified) across a wide range of benchmarks at comparable model sizes.

Current Vision-Language-Action (VLA) models for autonomous driving operate primarily through feedforward inference across the perception--reasoning--planning hierarchy. While modern architectures maintain temporal recurrence within the perceptual module, early perception remains blind to downstream reasoning and navigation goals, processing visual inputs agnostically without prioritizing cues informed by prior decisions. To bridge this gap, this paper introduces PRIME, a learned feedback mechanism that conditions the VLA perceptual queries on a novel Situational Memory. By aggregating latent representations of past perception, reasoning, navigation goals, and predicted behaviors across an L-step window via cross-attention, PRIME enables intent-driven perceptual attention at minimal computational cost, adding only a maximum of 29.7M parameters (0.41% of the 7.3B-parameter base model). Evaluated on the Bench2Drive closed-loop benchmark, PRIME achieves a state-of-the-art Driving Score of 82.47 (+4.73 over ORION) and a Success Rate of 60.00% (+5.38 percentage points), the highest reported Driving Score among published VLAs trained on Think2Drive demonstrations.

Users increasingly delegate work to autonomous AI agents, yet evaluations typically measure task completion rather than the values users prioritize. Using Value Sensitive Design, we analyzed, with LLM assistance, 73,093 first-person Reddit posts about using OpenClaw, each for its human value, agent aspect, value fulfillment, and user outcome. The 21 values form six value groups, including Autonomous, Dependable, and Affordable Operation, Bounded Reach, Reviewability, and Equitable Access. Relative to each aspect's corpus share, values clustered not at the agent's outputs but at the operating conditions users set around a run. Values were usually met where users described what the agent delivered, in five of six groups, and mostly unmet where users described supervising it, in all six groups. We conceptualize this pattern as value-sensitive delegation. Supporting human values requires attention not only to what an agent accomplishes, but to the conditions users set around delegation, including cost, access, and oversight.

Comprehensive video understanding is crucial for advancing artificial intelligence toward the intricate dynamics of the physical world. While recent advances in Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in video understanding, existing benchmarks remain confined to simple scene-level queries or global summaries that require only single-step inference. Real-world video understanding involves more challenging tasks that require multi-hop multimodal reasoning, and there is a critical absence of video benchmarks equipped to rigorously evaluate these agentic capabilities. To bridge this gap, we introduce AgentVidBench, a multi-hop video question answering benchmark focused on evaluating the spatial, temporal, and causal reasoning capabilities of MLLM agents. Beyond standard question-answer pairs, AgentVidBench provides step-by-step solution traces to support trajectory evaluation that assesses whether agents explicitly acquire the evidence needed to justify their answers. Experiments with 12 proprietary and open-source MLLMs show that single-turn performance remains limited on AgentVidBench, while integrating these models into state-of-the-art agentic workflows generally improves performance with respect to both accuracy and trajectory scores. We further present a simple yet effective agentic strategy that serves as a competitive baseline on AgentVidBench, establishing our benchmark as a holistic testbed for future research on agentic video understanding. Code and datasets are available at https://github.com/krafton-ai/agentvidbench and https://huggingface.co/datasets/agentvidbench/agentvidbench.

Multi-agent manipulation naturally produces multiple task-driven viewpoints: every arm carries a wrist camera and moves through the scene while acting. Yet these observations are typically underutilized, and active perception in manipulation is still often treated as requiring a dedicated sensing agent. We introduce MAAP (Multi-Agent Active Perception), in which every arm is dual-purpose: it executes manipulation actions and, through the wrist camera it carries, simultaneously serves as a moving viewpoint for the team. We pair this with RAIL (Role-Aware Imitation Learning), a controller that predicts each arm's current role alongside its action chunk and conditions action generation on it, representing role-dependent actions within one network. Across four simulated tasks, widening the perception regime lifts average success from 56.5% with a fixed camera to 62.5% with one active wrist view and 70.0% with all of them, while MAAP+RAIL reaches 79.2%. RAIL's additional gain is concentrated on the three-arm Microwave task, where success rises from 47% to 82% on identical multi-wrist inputs. On a dual-arm platform, MAAP+RAIL succeeds in 14 of 20 placement trials compared with 0 of 20 for fixed-view ACT. Collaborative manipulation can thus serve as an active perception mechanism in its own right.

Privacy leakage in LLM agents is commonly evaluated within individual components such as memory, retrieval, or tool-use pipelines, which makes it difficult to distinguish internal exposure from information that an external observer can actually recover. We present CIPL (Channel Inversion for Privacy Leakage), a channel-aware evaluation framework for black-box privacy leakage in LLM agents. CIPL represents a target through sensitive source, selection, assembly, execution, observation, and extraction stages and evaluates the transition from selected sensitive units to attacker-recoverable output under a shared protocol. Experiments across memory-based, retrieval-mediated, and tool-mediated targets, together with a BrowserUse live-agent case study, show that storage labels alone do not determine recoverability. Memory targets form a near-saturated reference case, retrieval-mediated leakage is frequently partial, and tool-mediated and live-agent leakage varies strongly with observation surface, prompt-to-channel alignment, retrieval depth, and provider behavior. A stratified semantic audit further identifies attacker-useful disclosures that canonical exact matching misses. CIPL therefore provides a common framework for comparing how internal sensitive dependence is realized as externally recoverable leakage across heterogeneous agent pipelines.

This study examines tool-augmented Large Language Model (LLM) systems for supporting Root Cause Analysis (RCA) of nightly test failures at Westermo Network Technologies AB. Nightly test executions produce heterogeneous test data and logs that practitioners currently inspect manually across multiple sources. We implemented an RCA workflow in single-agent and orchestrated multi-agent configurations, both with access to test metadata and logs. An exploratory industrial case study used two real failure scenarios. Six practitioners evaluated the scenario reports through a survey and focus group, and operational measurements were collected from 120 repeated executions. The evaluation covered practitioner-perceived correctness, reasoning quality, fix realism, clarity, usefulness, and trust, as well as cost, duration, and consistency. Neither configuration showed a consistent practitioner-perceived quality advantage across the two scenarios. The single agent system generated reports faster and at lower cost, making it the more practical baseline in this context. The potential benefits of agent architectures require further evaluation in more complex scenarios.

Field robotics missions often require physical samples to be returned to laboratories for analysis, making path planning inherently load-aware and order-dependent as accumulated samples increase payload and traversal energy costs. In single-robot Load-Aware Informative Path Planning (LIPP), this rigidly couples sensing with hauling: a solitary robot must transport every collected sample, forcing frequent depot returns that severely restrict its spatial coverage. Heterogeneous multi-robot teams can overcome this bottleneck by dividing labor---enabling high-precision samplers to collect while high-capacity carriers handle transport. However, this introduces a complex coordination challenge regarding when, where, what, and to whom handoffs should occur on top of the LIPP problem. To address this tightly coupled problem, we introduce Multi-Agent LIPP (MA-LIPP), which enables teams to cooperate through asynchronous "dead drops," allowing one robot to deposit samples for another to retrieve later without requiring synchronous rendezvous. We formulate MA-LIPP as an exact Mixed-Integer Quadratic Program (MIQP) alongside a scalable Pairwise Large-Neighborhood Search (LNS) heuristic for complex real-world applications. The heuristic matches exact optima in $95.5%$ of certified cases and reduces weighted posterior variance by $16.1$--$19.8%$ relative to a sequential baseline on larger instances of up to 12 robots, providing a robust framework for cooperative physical-sampling missions.

Reinforcement learning has substantially improved large language model (LLM) agents in verifiable domains, but remains difficult to apply to open-ended agent tasks, where solutions are diverse and reliable scalar rewards are hard to obtain. Recent pairwise evaluation methods alleviate reward discrimination collapse by replacing pointwise scoring with relative preferences. However, they still compress rich comparative feedback into a single trajectory-level reward, obscuring decisive intermediate steps and preventing successful behaviors from being consolidated into reusable skills. We propose ArenaFlow, a hierarchical credit propagation framework for open-ended agent reinforcement learning. ArenaFlow leverages tournament-based relative ranking to derive trajectory-level reward signals. Each comparison is further equipped with structured reflective evaluation, which reveals three types of supervision: pivotal success steps, reusable strategy skills, and usage attribution of retrieved skills. At the step level, ArenaFlow propagates trajectory-level advantages to high-confidence pivotal steps according to tournament survival depth, enabling more targeted optimization of local reasoning behaviors. At the skill level, ArenaFlow estimates skill utility from group-level usage attribution and maintains a global skill memory through utility-aware updating, pruning, and retrieval. The resulting high-utility skills further serve as policy priors for future exploration. Extensive experiments validate ArenaFlow's effectiveness on open-ended agent tasks.

Computer-use agents (CUAs) have advanced along two separate lines: graphical interaction and software development through code and the command line. Real digital work requires both, interleaved rather than stacked end to end. We study hybrid CUAs that autonomously decide when to explore an interface, implement software, and run and visually verify their artifacts. We introduce RecreationWorld, a five-platform framework built around recreation: given a running reference, an agent must discover its behavior and build a faithful implementation with no prescribed workflow. RecreationWorld provides reproducible environments on Ubuntu, macOS, Windows, Android, and Web, plus a unified harness with native GUI control and coding tools. The running reference serves as an oracle for hidden behavioral tests, providing execution-grounded rewards. We scale trajectory generation with high-quality open-source applications. Models trained on these trajectories improve across five out-of-distribution coding and hybrid computer-use benchmarks and more frequently verify their rendered outputs, providing evidence of transfer beyond recreation. For held-out evaluation, we introduce RecreationBench, comprising 250 diverse tasks across domains and platforms. Reference-grounded programmatic and visual assertions cover action-conditioned outcomes at multiple interaction depths; each is validated on the reference and by human reviewers before the suite is frozen for automatic scoring. GPT-6 Astra leads at 58.1% overall, but passes all programmatic tests on just 2.8% of tasks. Agents reproduce static interface structure more reliably than interactions and computed outputs, while generated applications remain smaller and more monolithic than their references. We release the benchmark, environments, and test suites.

Graph-enhanced multi-agent systems (G-MAS) coordinate large language model agents through communication graphs and role assignments, which determine how agents exchange information and divide responsibilities. However, final-score comparisons across systems combine differences in models, communication patterns, roles, and computation costs, making performance differences difficult to attribute to specific communication structures, role assignments, and information flows. To address this evaluation attribution problem, we introduce OpenMAS-GCom, a benchmark for diagnosing how these components affect G-MAS performance through controlled interventions. We represent systems through collaboration units, communication links, shared intermediate information, and execution rules. OpenMAS-GCom compares original systems with versions modified by changing one component while keeping tasks, models, prompts, and budget limits fixed. We rewire communication edges, remove specialist or critic agents, replace intermediate messages with incorrect content, and disable workers during execution. The benchmark evaluates 17 single-agent, ordinary multi-agent, and graph-enhanced configurations on 29 datasets across six domains. We add 400 G-MAS-Complex tasks requiring agents to combine information from multiple documents, resolve conflicting records, and return specified values with source identifiers. Experiments show larger mean losses after specialist removal than after critic removal, different performance degradation under incorrect messages and worker failures despite similar original scores, and different configurations achieving the highest accuracy and accuracy per token on G-MAS-Complex.

Agent models are frequently evaluated one decision at a time, where the model predicts the next action based on the gold interaction history, which is scored against a reference. We investigate whether improvement under this protocol is predictive of improved autonomous workflow execution. We study pre-SFT and supervised fine-tuned (SFT) Qwen3 models at 4B and 14B parameters and Gemma 3 models at 4B and 12B parameters on multi-turn customer-support workflows. We find that SFT consistently improves text-turn success, and that overall next-turn success increases for every model under gold-history evaluation. However, these improvements do not transfer to autonomous workflow execution. Tool-specific gains also vary across metrics and models. None of the four SFT models succeeds under holistic workflow evaluation, with strict trajectory completion reaching at most 10.4% workflow success. Our results show that next-turn evaluation is not a reliable proxy for workflow success, motivating separate reporting of text quality, local action correctness, tool execution, and end-to-end task completion.

LLM-based multi-agent systems generate collaboration traces that record how agents plan tasks, verify intermediate results, and repair failures. Reusing these procedures requires preserving an action's prerequisites and the outputs needed by subsequent agents. Our empirical studies show that grouping these dependencies into functional memory units improves their retention, while connecting units increases retrieval of the units and links jointly required by a task. The preferred combination of units also changes between instructions and checklists, even when each combination's content is fixed across formats. Updating choices from the outcomes of each combination and format pairing outperforms scoring combinations and formats separately. These findings motivate MACE, a memory-agent co-evolution framework that adapts memory organization and agent memory use through execution feedback. Its MemGoG structure represents functional units as subgraphs of related conditions, actions, and outputs, connecting them through support, conflict, and repair relations. MACE Loop selects task-relevant units and relations within a memory budget and provides each agent with instructions or checklists for its current operation. It records the selected units, presentation formats, agent outputs, and task outcomes to update unit scores and relations for retrieval and inform subsequent presentation choices. Across eight benchmarks, MACE outperforms ten baselines with an average score of 81.11%, compared with 78.97% for the strongest baseline, SAGE.

Nearly 90% of drugs entering clinical development ultimately fail, despite billions of dollars in investment. Pharmaceutical companies therefore rely on clinical development planning (CDP) and probability of technical and regulatory success assessment to anticipate development risks, yet these decisions remain labor-intensive and subjective, requiring experts across clinical science, statistics, regulatory affairs, and competitive intelligence to jointly acquire, synthesize, and reason over heterogeneous evidence. Here, we introduce TrialAtlas, a memory-augmented multi-agent research organization for CDP that mirrors this collaborative process by coordinating specialized agents for literature synthesis, competitive trial intelligence, regulatory precedent analysis, and integrated reasoning over trial design and development risk. TrialAtlas further learns from historical clinical trials and regulatory outcomes, including prior New Drug Applications (NDAs), to ground its decisions in accumulated development experience. To evaluate these capabilities in an authentic regulatory setting, we introduce TrialAtlasBench, constructed from 291 FDA Complete Response Letters and spanning three practical tasks: detecting trial design deficiencies, recommending actionable design improvements, and predicting technical and regulatory success. TrialAtlas achieves an F1 score of 50.0% for deficiency detection, outperforming the strongest baseline by 6.1 points, and reaches 85.3% balanced accuracy and 84.7% F1 for prediction of technical and regulatory success, improving over the best baselines by 6.7 points in balanced accuracy and 12.0 points in Cohen's kappa. In expert evaluation, 86.4% of TrialAtlas-generated concerns were judged valid, compared with 83.1% for OpenAI DeepResearch and 59.3% for Gemini DeepResearch.

Text-guided video editing modifies target content while preserving the appearance and temporal coherence of unedited regions. Training-based approaches provide strong control but demand substantial data and computation. Training-free methods fall into inversion-free and inversion-based paradigms. Inversion-free approaches avoid trajectory recovery, but their source-preserving guidance can limit editing strength and leave semantic changes incomplete. Inversion-based approaches recover a latent trajectory before regeneration, where approximation errors can accumulate and cause source-content drift and temporal inconsistency. We introduce Edit-VAR, the first training-free and inversion-free framework for text-guided video editing with a pretrained visual autoregressive video model. Edit-VAR directly encodes the source video into multi-scale discrete tokens and performs probability-guided conditional token replacement for source preservation. Attention-guided token-wise and scale-aware modulation selectively relaxes source constraints over edit-relevant positions and generation stages. Scale-Decoupled Generation, implemented as late-scale constraint release, regenerates motion-consistent details and reduces texture fragmentation. Residual-guided token pruning further exploits redundancy at the final two high-resolution scales to reduce inference cost. Extensive experiments and a blind user study demonstrate that Edit-VAR outperforms existing training-free video editing methods overall in editing fidelity, source preservation, temporal coherence, and inference efficiency.

Gating the value pathway of attention reportedly improves language model pretraining, and prior studies disagree on why. We argue and provide experimental evidence that such gates supply two different things that softmax attention lacks: abstention and noise filtering. The first is abstention, which allows an attention head to output nothing, bypassing the requirement that attention weights must sum to one. The second is noise filtering, which allows the value pathway of an attention head to suppress interference from superposed features in the residual stream. In our experiments in matched models from 10M to 350M parameters, we supply abstention through a learned per-head sink logit in the softmax and noise filtering through a gate on each value. We report three empirical findings. First, the benefit of abstention, measured as the reduction in validation loss relative to a matched baseline, declines as models grow, whereas the benefit of noise filtering increases with scale. In particular, abstention accounts for nearly all of the gain from gating at 10M and filtering for most of it at 350M. Second, the best model at every scale is the one with both primitives built in. Third, injecting controlled interference into the values a head reads confirms that the gate removes such interference, and reveals that each of the two gate forms we study has a characteristic blind spot. Supplying both primitives adds negligible parameters and remains compatible with the key-value cache.

Production LLM agents are evaluated repeatedly as they evolve, but full agent benchmarks are costly to rerun. We study efficient recurring evaluation for a production analytics agent serving tens of thousands of monthly active users and report first-hand deployment experience. Using 574 historical runs of the production benchmark, split chronologically into calibration and held-out periods, we compare random sampling, historical caching, fixed representative subsets, and IRT-based adaptive testing. The results show that multidimensional 2PL adaptive testing achieves the best overall score fidelity: executing 200 questions, 38.5% of a full run, yields 1.03 pp of MAE. We nevertheless deployed difficulty-stratified fixed subsets because of their operational simplicity, and show they transfer without recalibration to five other agent families and remain stable across calibration windows as short as one day. Drawing on this deployment experience, we report practical recommendations for recurring production-agent evaluation.

Vision-language-action (VLA) policies based on flow matching generate action chunks through repeated evaluations of an action expert. Increasing the number of integration steps raises inference cost, but does not necessarily improve closed-loop success. We propose Coda, which reallocates part of this integration budget to a single learned endpoint correction. A frozen policy first completes a few-step noise-to-action trajectory; a lightweight Transformer then predicts a demonstration-supervised residual using the candidate action, source noise, and shared observation-prefix cache. Only the corrector is trained. On 50 RoboTwin Easy tasks, five-step Coda improves success from 71.64% to 74.68% over the matched five-step baseline, while reducing forward latency by 30.2% relative to the default ten-step policy. A two-step configuration achieves 71.88% success with a 2.12$\times$ speedup. An independent 13-task control shows a 5.69-percentage-point gain at nearly equal latency, supporting correction as an effective alternative to additional integration. The same design also improves frozen official SmolVLA, raising two-step success from 60.8% to 69.4%. These results show that endpoint correction improves the quality-latency trade-off of frozen flow-matching policies.

Generative world models offer controllable and repeatable closed-loop simulation for end-to-end and vision-language-action driving policies, but production deployment exposes three unresolved requirements: faithfully reproducing a mixed fisheye-pinhole rig at native resolutions; reconciling causal, per-timestep interaction with long-horizon stability and low latency; and preserving scene identity when a location is revisited. We present ZYT-World, a single architecture that natively generates four fisheye views with field of view > 180° and three pinhole views. Projection-specific Plucker adapters encode camera geometry, ego-motion adaptive layer normalization provides global motion control, and a lightweight pixel-aligned layout conditions traffic participants and signals through instance-level boxes, headings and colors. Heterogeneous training combines full-rig geometric coverage with high-resolution detail. Teacher forcing, causal consistency distillation, self-rollout distribution matching distillation, and RigCritic transform a 40-step bidirectional teacher into a one-step, per-latent streaming generator, with RigCritic evaluating the seven-view rig jointly. A 19M-parameter variational autoencoder decoder (TinyVAE), W8A8 quantization, and our inference engine reduce decoding, backbone, and incremental-execution costs, respectively. Finally, cross-trajectory pairs derived from real captures train a plug-in implicit-memory module that preserves place-specific evidence. On the internal multi-view test set, the one-step model retains more than 90% of the teacher's PSNR and SSIM, while FID, FVD, and LPIPS stay within 11% of the teacher. Under the generator-only timing in Figure 2, it is 107.7 times faster than the 40-step bidirectional teacher. TinyVAE decodes 59.8 times faster than Wan. 30s rollouts and cross-trajectory revisits show the intended long-horizon and memory behavior.

Quantization offers a practical path to deploying diffusion models with reduced memory and computation, but aggressive compression can cause quantized outputs to deviate substantially from their full-precision counterparts. Sampling-stage correction methods seek to compensate for such deviations during sampling, but existing approaches rely primarily on local information and underexploit trajectory history, limiting their ability to correct errors that propagate across timesteps. In this work, we formulate sampling with a quantized denoiser as an online estimation problem, using the history of quantized denoiser outputs to recover the underlying full-precision outputs required by the sampler. We propose QuAKE, a Quantization-Aware Kalman Estimator that combines a smooth trajectory prior with a conditional Gaussian observation model. At each sampling step, QuAKE recursively updates the posterior over the output window in closed form and feeds its posterior mean to the sampler. QuAKE is a lightweight plug-and-play corrector that requires no modification to the quantized network and naturally supports arbitrary high-order multistep ODE samplers. Experiments across W4A4-quantized text-to-image diffusion models show that QuAKE consistently outperforms existing methods in reducing the distributional discrepancy from full-precision sampling.

Speculative decoding accelerates LLM inference by drafting multiple tokens in parallel, with tree-based methods further improving efficiency through hierarchical structures. Dynamic-tree methods such as EAGLE-3 perform well under greedy decoding via deterministic top-K expansion and global pruning. However, in stochastic decoding (T>0), this mechanism collapses the draft distribution into one-hot probabilities, causing a severe drop in acceptance rate. This creates a dilemma: dynamic-tree methods sacrifice stochastic sampling to preserve context-aware topology, while static-tree methods preserve stochastic sampling with context-agnostic structures. The issue arises because the same probability distribution is used for two conflicting tasks: constructing the tree and verifying tokens. This coupling makes direct injection of randomness challenging due to the resulting stochastic process. We resolve this by decoupling these roles: RheoSampling assigns a token sampled from the draft distribution a proxy probability for tree expansion and pruning alongside its true sampling probability for verification. Specifically, we inject a sampled token among the deterministic top-K slots and treat it with different probabilities during construction and verification, making RheoSampling the first dynamic-tree method with both context-aware top-K construction and stochastic sampling while maintaining losslessness. We establish the lossless guarantee through an equivalence-class analysis that compresses the stochastic tree space into tractable classes. An OT-based verification strategy and a sparse draft mechanism ensure that theoretical gains translate into practical efficiency. Experiments across LLMs and benchmarks demonstrate improvements in acceptance rate and speedup over state-of-the-art dynamic tree methods. This framework may provide a template for analyzing stochastic tree structures.

Spatial NPUs such as AMD XDNA place compute tiles beside small local memories and leave data movement between them to software. Mapping a multi-stage workload onto such a device is largely a question of where the intermediate tensors live. We report what we learned making those choices for FlashAttention with the open-source IRON and MLIR-AIR flows. We compare four reference designs on XDNA 1 and XDNA 2: one runs each operator separately, two stream between operators on chip, and one fuses all three attention stages into a single kernel. The fused kernel holds the $\boldsymbol{QK}^{\mathsf T}$ scores in compute-tile local memory and reduces partial results over the cascade interconnect, so the scores never return to shared MemTile memory. On XDNA 2, it reaches 3.62 TFLOP/s over complete end-to-end execution, twice the IRON design, with 5.3 to 7.2 times the energy efficiency of the integrated GPU on the same chip at 2K tokens and above. It covers twelve LLM configurations, from BERT to DeepSeek, up to 128K tokens. Roofline analysis at each memory level explains this result and shows when to stop. XDNA 1 has lower ridge points, so streaming on chip already reaches the compute-bound regime: the same fusion that doubles throughput on XDNA 2 is nearly wasted on XDNA 1. Comparing a mapping's operational intensity against each level's ridge point predicts which case applies before writing any code. Fuse until the mapping clears that ridge point, then stop. We release the reference designs as maintained open source.

Large language models (LLMs) are moving onto mobile devices for increasingly diverse workloads over text, images, video, and audio. These applications often require long contexts, making the Key-Value (KV) cache a dominant memory bottleneck because it grows linearly with sequence length and is accessed at every decoding step. Prior work reduces KV-cache footprint through low-rank compression, token eviction, or flash offloading, but the resulting reconstruction overhead, irreversible token loss, or I/O stalls can offset the benefit of saving memory. We present TierKV, a mobile LLM inference framework built on Predictive Multi-Tier Cache Optimization (PMCO). Before decoding starts, PMCO predicts future cache demand from prefill hidden states and jointly assigns tokens to exact, low-rank, and flash-offloaded tiers under the device memory and accuracy budgets. This formulation retains access to the full context, removes the circular dependency of reactive eviction, and admits a closed-form solver that selects tier boundaries and per-layer ranks at runtime. Across eight text, vision, and audio models on three mobile SoCs, TierKV improves prefill throughput by up to 17.6x over existing mobile LLM frameworks, reduces RAM-resident KV cache by 12.5-34%, thereby enabling substantially longer contexts under the same memory budget, while incurring only minor accuracy degradation.

Running large language models (LLMs) locally continues to be limited by restrictions of compute and memory on consumer hardware. The popular acceleration technologies, such as quantization, speculative decoding, and adaptive inferencing, offer substantial speed boosts but usually necessitate retraining, per architecture tuning, or draft models. SpecQuant is a trainingfree framework, that combines speculative decoding with multiparent quantization to perform adaptive, efficient inference of LLMs. SpecQuant derives multiple quantized variants (INT4, FP8, FP16) from a shared base model, and dynamically routes queries based on predicted complexity; lightweight variants are used for simple or factual tasks, and full-precision models are used for complex reasoning tasks or long-context inputs. The shared-weight design of SpecQuant ensures sufficient token acceptance for speculative decoding without compatibility issues using separate draft parent models. We evaluate SpecQuant on Qwen2.5 based models on the MMLU, AlpacaEval, and GSM8K datasets, or benchmarks, demonstrating 35-43% speedups without degrading accuracy greater than 2%, substantial within the LLM community. SpecQuant enables practical on-device LLM deployment across diverse hardware without special infrastructure or expertise.

The growing elderly population gives rise to an urgent need for intelligent support systems, particularly for individuals with Mild Cognitive Impairment (MCI). This paper presents PV-Care, a proactive AI-driven assistance scheme that integrates wearable electroencephalogram (EEG) sensing with visual environmental perception to provide real-time, context-aware voice assistance for MCI users. Unlike traditional assistant systems that passively wait for user commands, PV-Care actively initiates helpful interactions based on the user's detected brain states, including Learning, Memory Recall, and Resting, using a novel deep neural architecture named Spatial and Frequency Refinement Network (SFR-Net). By combining EEG-based cognitive-state recognition with AI-based visual analysis, PV-Care generates structured "4W-UT" prompts to guide the output of large language models (LLMs). Simulation results and user studies validate the high accuracy of the proposed SFR-Net and the effectiveness of PV-Care's context-aware assistance. These results indicate that PV-Care is a feasible and promising solution for MCI caring.

The success of Large Audio Language Models has driven the development of massive multimodal networks exceeding billions of parameters. However, the demand for privacy-preserving, low-latency processing has shifted focus toward Small Audio Language Models (SALMs) capable of on-device execution. In this paper, we introduce Samsone, a family of SALMs designed for edge computing. Our core model, Samsone-134M, establishes a new state-of-the-art for its size class across multiple benchmarks. We further explore the scaling laws of SALMs by introducing Samsone-99M and Samsone-356M. Despite their compact footprint, the Samsone family delivers performance competitive with models orders of magnitude larger. To foster open research and reproducibility, we train Samsone on publicly available data. We release the training code, model weights, mobile-optimized checkpoints and provide an open-source Android application to demonstrate real-time on-device inference of Samsone.

Best-of-$n$ (BoN) sampling is a simple yet effective inference-time alignment method, but hard maximization provides only coarse control over the trade-off between reward and distribution shift. Soft Best-of-$n$ (Verdun et al. 2025) provides smoother control and converges to the optimal distribution associated with KL-regularized reward maximization. In this paper, we introduce ExpBoN, an alternative soft BoN method based on the exponential-noise report-noisy-max mechanism. It admits an exact finite-$n$ decomposition, which yields exponentially fast convergence in total variation, expected reward, and both directions of KL divergence. We provide comprehensive theoretical analyses of its convergence and regret behavior. We further integrate ExpBoN into the guided speculative inference (GSI) framework (Geuter, Mroueh, and AlvarezMelis 2025), resulting in ExpGSI, for efficient reward-guided LLM alignment. ExpGSI yields substantial reductions in computational cost while maintaining comparable accuracy. Experiments on MATH500, MMLU-STEM, and Minerva Math with the Qwen2.5-Math and Qwen3 model families show that ExpGSI reduces estimated computation by $14%$-$39%$ across candidate budgets for Qwen2.5-Math and by up to $45%$ at $n=16$ for Qwen3. Overall, our results provide a theoretical and algorithmic foundation for exponential-noise BoN and efficient test-time LLM alignment.

Large language models can hold knowledge they do not report. A model may sandbag on a capability evaluation, or answer against what it internally knows, and its outputs alone cannot tell whether it is hiding an answer or simply does not have one. We borrow the Concealed Information Test, a forensic method that identifies guilty knowledge by presenting a suspect with the true detail among plausible decoys and measuring a stronger response to the item they recognize. Our method, Probe of Internal Recognition (PIR), does the same inside a model. It presents a question with its candidate answers and reads, from the model's internal states, which candidate the model recognizes as correct. PIR is reference-free, needing no honest reference model and no labeled truth corpus. Across eight models from five families (Gemma, Qwen, Llama, Mistral, and Phi), PIR recovers the recognized answer at 0.70 to 0.87 balanced accuracy, well above the 0.28 to 0.40 unknown-item baseline and the 0.25 chance rate. It stays readable across every form of concealment we test, from prompted deception and trained sandbagging to external password-locked and circuit-broken checkpoints, with recognition between 0.85 and 0.93. When the model hides a known answer, recognition stays high. When unlearning removes the knowledge, recognition drops to the level of a question the model never knew. PIR therefore separates a model that will not answer from one that cannot, which supports sandbagging audits and unlearning verification. The signal is causal, adds information beyond black-box behavioral cues, and extends from multiple-choice questions to free-form generation.