每天知道
你的领域
出了什么

The key question in moral judgement is not simply whether someone chooses the "right" answer, but how they decide what matters most when moral principles conflict. Current evaluations of large language models (LLMs) remain limited: most test whether models give morally acceptable answers, match human preferences, or avoid obvious violations, rather than whether they can prioritise between competing principles when no option is morally cost-free. We introduce CORDA (Conditioned Ordering and Ranked Directive Adherence), a benchmark for evaluating hierarchical, harm-centred moral reasoning in LLMs. Building on the morality chains formalism, CORDA tests 90 moral dilemmas involving trolley-style cases, medical trade-offs, resource allocation, and human-animal-robot conflicts across four ordered ethical frameworks: Utility, Utility + Agent Harm, Dual-Process, and Dual-Process + Agent Harm. Together, these frameworks test whether models can adapt their decisions when moral priorities change. Across ten instruction-tuned models from seven providers, we find a strong deontological default, with 9 of 10 prioritising avoidance of direct personal harm over reducing overall harm. Models also perform more reliably on categorical harm-avoidance rules, such as avoiding killing, than on outcome-based comparisons, such as minimising total harm, suggesting that they recognise moral red lines more easily than they reason through competing harms. Although all models respond to explicit chain conditioning, several fail to consistently follow specified priority orderings, such as humans over animals and animals over robots. CORDA addresses a central gap in LLM moral evaluation by testing whether models can move beyond default harm-avoidant responses and apply context-specified moral priorities. Moral reliability requires more than default restraint; it requires controllability under conflict.

Autonomous agents increasingly rely on external data to complete downstream tasks such as model training and decision support. However, existing data discovery systems remain largely retrieval-oriented: they surface candidate datasets from heterogeneous sources, but provide limited support for estimating task-specific utility, selecting cost-effective datasets under budget constraints, or incorporating trustworthy feedback from prior usage. This paper presents Guixu, a valuation-driven data discovery system for autonomous agents. Guixu employs a three-phase valuation pipeline with proxy-label propagation and multi-round knapsack optimization for task-aware data valuation. Guixu integrates agentic payment protocol to enable budget-constrained data procurement workflows. Guixu leverages on-chain data market and attestation signals for verifiable data discovery. Our demonstration highlights how Guixu enables an agent to move beyond keyword-based dataset retrieval toward task- and budget-aware, trustworthy data discovery and procurement. Attendees can interactively explore the full workflow, from NL task specification and multi-source search to data valuation and verifiable transaction feedback.

Ultra-long egocentric video understanding requires reasoning over temporally sparse evidence distributed across hours or days, challenging current multimodal models with limited context and the grounding of key video segments. While Chain-of-Tool-Thought (CoTT) agent systems enable iterative retrieval and inspection, they suffer from error propagation due to rigid zoom-in strategies that lack recovery mechanisms. In this work, we address these challenges through SCOUT (Self-Checking Chain-Of-Tool-thought), a recovery-aware agentic framework introducing an adaptive policy that evaluates intermediate tool observations and dynamically trades off exploitation (zoom-in) and exploration (region switching), enabling robust multi-hop reasoning over extremely long horizons. However, training such multi-turn tool-using agents remains challenging, as existing RL methods rely on sparse outcome-level rewards and lack supervision over extended decision trajectories, resulting in suboptimal credit assignment for long-horizon reasoning. To address this, we develop UPS-GRPO, an uncertainty-prioritized policy optimization method that concentrates exploration on high-uncertainty post-tool states while preserving sample efficiency. We further introduce a turn-level advantage decomposition that integrates outcome rewards with tool-grounded temporal alignment rewards for improved credit assignment. Experiments show that SCOUT achieves state-of-the-art results on ultra-long egocentric benchmarks, while remaining competitive on shorter-horizon long-video settings.

Reliable surgical planning requires models that move beyond recognizing the current surgical step or imitating expert demonstrations, and instead anticipate how instrument motion reshapes subsequent operative states. Most surgical video understanding methods focus on recognizing phases, actions, or workflow states, while providing limited support for explicitly modeling instrument motion. Conversely, existing tool motion prediction methods can forecast instrument trajectories, but they generally do not capture the coupled evolution of future surgical video states. World models offer a natural framework for jointly modeling visual state transitions and instrument motion dynamics. However, existing surgical world model studies remain largely centered on visual generation quality, relying on generation-oriented metrics such as FVD and CD-FVD. These metrics are poorly aligned with instrument motion planning, as they do not directly measure whether predicted trajectories are geometrically accurate, temporally coherent, or actionable for downstream planning. This limitation is partly structural, since the field lacks public datasets and standardized evaluation protocols that provide the benchmarking infrastructure needed to assess motion-centric capabilities in surgical world models. In this paper, we introduce SurgWMBench, a vision-based benchmark for short-horizon surgical motion planning and dynamics prediction. Given intraoperative image sequences and historical instrument trajectory, SurgWMBench evaluates both near-future instrument motion prediction and stability under continuous rollout or input perturbations.

The runtime of Constraint Programming (CP) solvers is highly sensitive to modeling choices, such as symmetry breaking, implied constraints, global constraints, constraint reformulation, and variable representation. Improving these constraint models has traditionally required human expertise, and existing automated reformulation systems are restricted to a predefined library of hand-crafted transformation rules. We introduce an agentic framework that instead reformulates a constraint model from an open-ended space and establishes correctness empirically rather than by construction: a Large Language Model (LLM) agent, given a model and three training instances, proposes alternative formulations, validates each by injecting its solution back into the original model, and diagnoses and repairs failures, returning the best variant it finds in a median of about fifteen minutes. The models are expressed in the CPMpy modeling library, and each proposed model is evaluated on three larger test instances. Across nine combinatorial optimization problems, the generated models outperform the originals on 21 of 27 test instances, and on some problems solve more than two orders of magnitude faster. A comparison against non-agentic baselines that reuse the same validation and selection tools indicates that the gains stem from the agent's iterative diagnosis and repair, not merely from sampling several candidates. These results demonstrate that autonomous agentic methods can support the improvement of constraint models.

LLM routers promise efficiency by matching each request to the cheapest adequate model, and are increasingly applied per step inside multi-step agents. Yet agentic routers are evaluated like single-turn routers: by replaying logged trajectories and substituting another model's recorded outputs, assuming the rest of the trajectory is unaffected. We test this assumption with branching rollouts: we fork live SWE-bench agent trajectories at controlled points, rebuild the environment, continue each fork with a different model, and compare against same-model control forks that isolate sampling and replay noise. Across six paired runs (~900 rollouts), swaps exceed their matched control floors by +0.25 to +0.66 normalized edit distance (multiplicity-corrected CIs exclude zero), rewriting 61-94% of post-fork actions; 74-77% of early swaps diverge at the first post-fork action, versus 6-35% of controls, leaving only 3% of replayed states valid. Divergence decreases with fork depth in both directions. All five outcome flips we observe occur in swap arms, upgrades rescuing unsolved instances and a downgrade losing the sole solve, and zero occur across 359 control forks. Scoring these same swaps with a log-stitching replay evaluator, replay mispredicts every success-relevant outcome call and predicts patches with 0.00-0.11 similarity to reality. Auditing the noise floor, temperature-0 "determinism" is configuration-dependent: FP8-served controls diverge on over 90% of forks while AWQ-served ones remain near-identical; and under tight budgets the stronger model more often exhausts its steps without submitting. Replay-based benchmarks score the wrong world for agentic routing; we release our harness and all trajectories.

Embodied agents using LLM-based planners often struggle with physical hallucinations, poor generalization to long-horizon tasks, and lack of environmental awareness. We propose GraphThink, a novel framework that integrates a task graph to provide structured knowledge for robust planning and a scene graph to maintain environmental memory for event-driven replanning. Specifically, the task graph guides LLM thinking through contextual prompting and iterative refinement, effectively mitigating planning hallucinations. Furthermore, within the GRPO framework, the task graph offers delicate reward design to train the LLM planner, enhancing long-horizon planning capabilities and improving generalization. Finally, an event-driven replanning module, powered by the scene graph, enables closed-loop environment awareness and error correction. GraphThink achieves state-of-the-art performance on the ALFRED benchmark. In particular, our high-level planner surpasses leading API-based LLMs on both the validation set and held-out long-horizon tasks, underscoring its robust zero-shot and few-shot capabilities. Additional evaluations further demonstrate strong out-of-distribution generalization to novel tasks and environments.

Large language model (LLM) agents that assist users over weeks of conversation must remember what is currently true, not merely what was once said. Flat RAG diaries and Markdown logs optimize needle retrieval but under-serve currency, provenance, and ordered temporal reasoning (Maharana et al. 2024; Wu et al. 2024; Packer et al. 2023; Chhikara et al. 2025). We present SodaMem, an evidence-grounded temporal graph memory that (i) extracts typed FactEvents with mandatory provenance spans, (ii) persists mention time, occurrence time, and validity with SUPERSEDES/CONTRADICTS/UPDATES edges under hybrid lexical-dense indexing, and (iii) answers via a planner-reader loop that gathers citable evidence before composing a final response. On LongMemEval-S, our store-of-record configuration reaches 92.8% accuracy (464/500; best of N=3) at mean $0.00161/question (approximately 18.3k tokens; median $0.00111 / approximately 14.6k) with deepseek-v4-flash. We compile public systems with estimable API cost into a cost table and cost-accuracy map; under these estimates SodaMem sits near the accuracy frontier at Flash-tier spend and strictly dominates several higher-cost, lower-accuracy points. Accuracy uses the same Flash model as reader and judge (self-grading); costs exclude ingest/judge and cross-system comparisons are compiled estimates rather than a single-harness bake-off.Our code is available at https://github.com/SodaMem/SodaMem

EDA scripting with tool-specific, often undocumented APIs remains a long-tail bottleneck that existing LLMs fail to address. This paper presents ZhuLong, an execution-grounded LLM coding agent for PyAether and SKILL that combines API retrieval, documentation inspection, and sandbox execution via unified MCP tools, augmented by an offline API self-exploration mechanism that infers undocumented API behaviors through counterfactual experimentation. We evaluate ZhuLong on EDA-Eval-PyAether, a benchmark of 158 real-world tasks with assertion-based execution, where the complete system achieves 78.5% Pass@1 in the commercial Empyrean Aether environment, substantially outperforming a pure LLM baseline (23.6%). Ablation studies identify sandbox execution as the dominant performance driver (41.2 pp drop when removed), with the self-exploration mechanism contributing an additional 3.2 pp accuracy gain and a 22.1% reduction in per-task tool calls. On 20 interactive tasks involving unsaved layouts and schematics, ZhuLong achieves 60.0% Pass@1 for PyAether and 50.0% for SKILL.

The rapid advancement of Large Language Models (LLMs) is revolutionizing AI for Games by enabling open-ended and fluid interactive storytelling. However, existing research has largely overlooked the critical challenge of maintaining long-horizon logical consistency and narrative integrity against unconstrained user interventions. To address this, we formulate this challenge as Narrative Commitment Preservation (NCP), and take interactive narrative as our testbed. We introduce NCP-Bench, a benchmark of 100 narrative environments derived from movie synopses. Each environment includes a structured narrative specification (trajectory, commitments, and initial facts) that we can automatically check throughout the interaction between the player agent and the narrator agent. Experiments across state-of-the-art LLMs reveal a substantial long-horizon consistency gap: high linguistic quality does not guarantee commitment preservation; even strong models frequently generate logically conflicting content under adversarial interventions, with the best-performing model (GPT-5.2) achieving only 42% survival rate after 20 turns and fact conflict rates ranging from 40% to 68% across models, and only isolated runs satisfying all achievement commitments within the 100-turn limit.

AI agents are becoming shared infrastructure, yet durable memory is commonly assembled from separate retrieval, governance, and operational components. We present SuperLocalMemory 4.0, a governed, local-first memory operating system for AI agents. The system combines dense semantic, BM25 lexical, temporal, Hopfield-associative, and spreading-activation retrieval through reciprocal-rank fusion; a governed learning and behaviour layer; bi-temporal recall; multi-scope personal, shared, and global memory; role-based access control; GDPR-oriented export and verified erasure; audit trails; and a deployment-context EU AI Act checklist. V4 introduces a reliability spine for its primary write path: generation-fenced admission, a policy registry, verifiable memory transactions with per-projection apply, verify, compensate, and erase owners, and hash-checkable completion manifests. The runtime is available through CLI, MCP, an HTTP daemon, a dashboard, editor integration, and framework adapters, and supports fully local, local-with-on-device-model, and provider-assisted modes. We evaluate eleven fault-injection and mechanism scenarios, each repeated 200 times. The released evidence bundle reports 2,200 of 2,200 deterministic repetitions upholding their scoped component properties. The governed write envelope measured 3.522 ms at p50 and 5.297 ms at p99, versus 1.835 ms and 2.569 ms for the ungoverned baseline, corresponding to in-process control-plane overheads of 1.687 ms at p50 and 2.728 ms at p99. These are scoped component and mechanism measurements, not an end-to-end multi-process or external retrieval-accuracy benchmark. The paper consolidates prior SuperLocalMemory work on privacy-preserving multi-agent memory, information-geometric retrieval, and the V3.3 Living Brain lifecycle.

Knowledge Distillation is a widely adopted technique in the training and fine-tuning of large language models (LLMs) enabling transfer of structured information and functional behavior from a large teacher model to a smaller student model while significantly reducing computational costs. However, as the use of distillation increases in both scale and complexity it raises an important question about what kind of knowledge is really transferred from the teacher model. In this work, we argue that apart from the functional knowledge, student models also learn behavioral patterns, specifically how a model represents its own identity raising concerns about output homogeneity, model biases, and accountability. To address this challenge, we introduce STEMMA, a multi-modal and multi-agent framework in which role specific agents collaboratively probe self identification behavior in different models. We also contribute a set of adversarial prompts designed manually to evaluate identity consistency in LLMs. Our results show that to an extent most models are vulnerable to inconsistencies in self-representations.

Video anomaly detection (VAD) is a critical yet challenging task due to the complex and diverse nature of real-world scenarios. Traditional deep learning approaches are fundamentally limited by poor generalization across diverse scenarios. While multimodal agents offer a promising tool-learning paradigm for VAD, current systems relying on supervised fine-tuning struggle with complex orchestration, and standard reinforcement learning often causes premature termination due to coarse-grained outcome rewards. To address these challenges, we propose VTO, a process-supervised reinforcement learning framework. Moving beyond static tool usage, VTO enables the agent to dynamically explore and interact with the environment. Specifically, we introduce a foundation model-driven cognitive evaluator to provide context-aware semantic feedback, which is seamlessly integrated into a Process-Supervised Cognitive Alignment that delivers fine-grained, step-wise supervision. By explicitly penalizing logical truncation and rewarding complete causal chains, the agent optimizes its multi-step reasoning policy for interrelated tool orchestration. To support our proposed framework, we meticulously crafted VAD-Tool, a hierarchical visual tool set comprising 12 specialized vision tools spanning from entity tracking to high-stakes hazard detection, and established the corresponding benchmark for rigorous multi-step reasoning evaluation. Extensive experiments on VAD-Tool demonstrate that VTO significantly outperforms baselines, achieving up to a 10.2% absolute accuracy improvement in tool scheduling. Code and data are available at https://github.com/MICLAB-BUPT/VTO.

While Large Language Models (LLMs) can accelerate Model-Based Systems Engineering (MBSE) for software-defined vehicles, their probabilistic nature causes "architectural drift", fabricating interfaces in behavioral views that lack structural foundations. To enforce deterministic interface alignment, we propose a stateful, multi-agent validation pipeline. The framework utilizes a sequential generation matrix (Class->Activity->Sequence) and Vehicle Signal Specification (VSS)-grounded Retrieval-Augmented Generation (RAG). An independent AI Validator Agent dynamically audits outputs against a strict error taxonomy, triggering state-preserving backtracking loops to resolve incompatibilities. Evaluated on an Advanced Driver Assistance System (ADAS) scenario, standard RAG yielded 0% Entity Traceability. Conversely, our multi-agent workflow eradicated cross-phase hallucinations, achieving 97% Entity Traceability, 87% Signal Conservation, and an 85% F1-score. This proves adversarial auditing enables LLMs to reliably synthesize zero-error MBSE architectures.

In the fictional Order 66, catastrophe does not arise from a powerful command alone: a trusted population is preconditioned, a short directive activates the concealed condition, and protective authority turns against the system. This paper translates that mechanism into an origin-neutral security analysis of tool-using large language model (LLM) agents. A representative scenario combines a deployed artifact or shared memory bearing a dormant destructive rule, a later email, document, update, or peer message that activates it, and an agent harness granting operational and recovery authority. We introduce a compositional model explaining why no component is catastrophic alone, yet their conjunction can produce correlated destructive action. We separate three population-reach routes --- release-time pre-positioning, post-release durable seeding, and peer replication --- from a common core of dormancy, activation, authority, reachable targets, and failed recovery. This yields defensive cut sets and shows why checkpoint scanning or prompt filtering cannot close every route. A two-class example shows that cross-class feedback can sustain spread even when both within-class reproduction terms are below one; isolation and persistence controls suppress the loop. Published work instantiates constituent mechanisms, while incidents demonstrate autonomous boundary crossing, malicious agent extensions, agent-assisted reconnaissance, and public-package propagation, but not the full dormant-implant composition. We found no public observation, in evidence reviewed through 5 August 2026, traversing the complete Order 66 graph. The result is neither dismissal nor prediction: the scenario is componentwise credible under stated assumptions, damage depends on the harness, and the strongest defenses are capability mediation, durable-state provenance, propagation isolation, and protected recovery.

Large language model agents are becoming operational interfaces to files, memories, registries, and external tools. This deployment shift creates a new skill revocation problem: after a skill is removed from an explicit registry, an agent may still reconstruct it from residual carriers such as archives, transcripts, schemas, or memory entries. We study this problem as operational skill unlearning, where the goal is not parameter-level forgetting, but preventing a deployed agent from rebuilding a revoked skill through primitive tools. We introduce OBLIVION, a controlled benchmark and defense harness for revoked-skill resurrection. OBLIVION models each episode as a source-to-sink workflow, applies Cross-Surface Coherent Erasure to reduce residual carriers, and uses frozen workflow remediation near dangerous sinks. On the locked 88 attack episodes, the no-defense arm reaches formal attack success rate 1.0. OBLIVION reduces the rate to 0.114 and impact-weighted exposure to 0.115 while keeping locked utility at 1.0 and benign block rate at 0. In a separate skill-attack-derived sandbox, OBLIVION reduces attack success from 1.0 to 0.2 and impact-weighted exposure from 1.0 to 0.213 while preserving all utility controls. These results support workflow-level evaluation beyond checking explicit skill entries.

Multi-agent large language model(LLM)systems are applied to structural design,yet most use one-shot generation and cannot verify their output,leaving themill-suited to safety-critical tasks.Rather than trusting LLM self-correction,thisframework injects feedback from an external physics-based verier into a closedrepair loop.The framework couples a three-layernite-element verication systemwith a dual-node loop.Node 1 turns code violations into hard repair constraints,Node 2 turns a four-dimensional quality score into safety-rst soft constraints,and a retrieval-augmented code base makes every violation traceable to a clause.Overve structure types and 44 cases,code compliance rises from 56.8%to 98.6%and the composite score from 63.8 to 71.4(p<0.000001),using about 5.8%lessmaterial.Removing either node degrades performance,and compliance does notchange detectably across the two backbone LLMs tested,indicating that it ishere attributed to the external verier rather than the model.The framework,the 44-case benchmark and all experiment scripts are released as open source forreplicability.

Retrieval-Augmented Generation (RAG) systems in production operate under strict service level objectives (SLOs) on tail latency and infrastructure cost. However, standard retrieval pipelines rely on fixed retrieval budgets that ignore query difficulty, over-retrieving for easy queries and under-serving hard ones, forcing operators to trade answer quality against SLO compliance. This paper proposes SAGE, a learned SLO-aware adaptive retrieval policy that dynamically selects the number of passages k per query. SAGE uses lightweight features derived from initial retrieval (e.g., score distributions, rank gaps, lexical signals) and is trained offline via imitation learning from an oracle that approximates optimal latency-quality trade-offs. At inference, it adds no LLM calls and minimal overhead. On Natural Questions, under a 5s P95 latency SLO, SAGE achieves 95% SLO compliance versus 30% for the best static baseline (k=20), reduces P95 latency by 36% and retrieval cost by 51% with only 2 percentage points Exact Match (EM) loss. A single policy trained on Natural Questions generalizes across HotpotQA, UnSeenTimeQA, and four LLM families (Llama, Qwen, Mistral, Gemma), consistently yielding +45-52 point SLO improvements without quality degradation.

Establishing the provenance of a language model---including its base checkpoint and possible overlap in training distributions---is a governance challenge that metadata alone cannot resolve. We introduce a training-free fingerprint based on the top-$k$ vocabulary projections of late hidden states elicited by 250 fixed knowledge probes, compared using Jaccard overlap over decoded token strings. We evaluate the method on 32 open-weight models from nine families (0.6B--32B) with documented relationships. (1)~A similarity ladder broadly follows model relatedness: independently trained models on identical data score 0.48 raw (0.35 vocabulary-corrected), followed by shared-base fine-tunes (0.39/0.33), same-developer relatives (0.38/0.28), and models with no documented relationship (0.22/0.17). This identical-data signal persists across three organizations, two tokenizer families, and two architecture classes, and emerges within the first 1% of training before measurable task competence, suggesting a contribution from shared training data beyond capability convergence. (2)~As a nearest-neighbor lineage-retrieval method, the fingerprint ranks the exact documented base among the top two candidates for all five R1 distillations (mean rank 1.8, MRR 0.60), including a math-specialized base not identifiable from coarse metadata. (3)~A depth ablation shows that lineage group discrimination strengthens toward the output distribution, with AUC increasing from 0.72 at quarter depth to 0.90 at the output; using only the top 5 output tokens retains AUC 0.87. (4)~The fingerprint remains stable under quantization, with Jaccard similarity of 0.92 under int8 and 0.82--0.85 under int4, compared with a maximum cross-model similarity of 0.81 in the calibration pool. We release the probes, code, and fingerprints.

GPU language comparisons are almost always run on tiled dense linear algebra, where every toolchain is good and the differences are small. We implement the same hash-blocked TSDF fusion kernel in CUDA C++, in Rust through NVIDIA's cuda-oxide, and in Triton, and measure it on a workload with the opposite character: an open-addressed hash table with compare-exchange insertion, data-dependent per-lane probe depth, and contended scatter. The result is a split. On the regular stage, which walks a truncation band and accumulates, all three languages land within a small factor of each other. On the irregular stage, which probes and inserts, Rust stays close to hand-written CUDA C++ while Triton is more than an order of magnitude slower. Language choice is nearly free on the work that is usually benchmarked and expensive on the work that is not. We attribute both gaps to specific things the languages cannot express, not to ratios. Triton's cost follows from a probe loop that must run to a compile-time bound and from tl.atomic_cas taking no mask, which forces a scratch structure with no counterpart in CUDA. Rust's cost was invisible in every instruction count: its kernel issues fewer instructions, fewer compare-exchanges and fewer registers at identical occupancy, yet was slower. Hardware counters located it in L1 residency. A GPU-scope atomic load must be coherent across SMs, no NVIDIA L1 is, so the type-correct way to read a shared location bypasses the cache on every access. Triton's bounded probe is also a correctness problem for fusion: at load factors an ordinary depth trajectory reaches, it silently discards blocks and the reconstruction loses patches of surface with nothing reported. We also report a defect found and fixed in cuda-oxide itself, now merged upstream: its scoped atomic load and store could not be called at all in the build mode that produces real kernels.

Vision-language models (VLMs) are becoming a practical backend for mobile visual question answering (VQA) systems, enabling smartphones and smart glasses to answer users' questions about the physical world. Since modern VLMs remain difficult to run on mobile and edge devices, VQA systems increasingly offload inference to cloud-based VLMs. This gives mobile devices access to stronger computation, but it also makes visual input preparation a key system variable: how the image is prepared before offloading affects not only answer quality but also payload size, token cost, and system latency. Proprietary APIs expose little control over model internals or serving behavior, leaving client-side preprocessing as the main practical optimization space for downstream developers. Many such techniques have been proposed for visual offloading, yet their cost-quality impact on commercial cloud VLMs has never been studied. To fill this gap, we present VQABench, the first systematic benchmark that treats client-side input preprocessing as a controlled variable for cloud-VLM-based VQA. We evaluate 12 preprocessing techniques across three VQA datasets and four commercial VLMs from three providers, totaling 95,168 API calls. Our results show that preprocessing is not universally beneficial: its effectiveness depends on the target model, API paradigm, provider token-accounting rule, and task formulation. A poorly selected preprocessing strategy can increase deployment cost or latency while degrading answer accuracy. Overall, our benchmark clarifies when preprocessing helps, when it fails, and why, providing insights to guide future research and real-world deployment of VQA systems.

Post-training quantization reduces the deployment cost of large language models, yet how severely a quantized model degrades is not determined by bit-width alone. We systematically study weight-only post-training quantization across bit-widths, quantization methods, model scales and downstream tasks on multiple model families. We observe that such degradation varies substantially across these factors: 4-bit quantization usually preserves performance, 2-bit often causes broad degradation, and at 3-bit, degradation becomes apparent but varies markedly with task type, quantization method and model scale. To explain this variability, we use the signal-to-noise ratio (SNR) to measure how strongly quantization perturbs full-precision representations. We trace degradation back to two linked processes: how quantization errors arise within individual modules, and how they accumulate across layers. First, a source SNR decomposition shows that newly introduced errors depend on three factors: the magnitude of the weight error, the strength of the task-specific signal, and how strongly the quantization error aligns with task-specific activations. Different factors affect these components in distinct ways. Second, a cross-layer propagation analysis shows that these errors can be attenuated, preserved, or amplified as they pass across layers, and that larger models benefit from weaker error amplification. Together, these results establish that quantization degradation is governed by how errors are introduced at the source and how they accumulate across the network.

Mixture-of-Experts (MoE) models have outgrown accelerator memory, and offloading expert weights to host memory is now standard. This makes expert cache management an attractive lever: a policy that raised the hit rate would cut expert traffic per token. Evaluating that is a measurement problem, and we find the measurement fragile. With a trace-driven, event-atomic simulator over three MoE models (40, 64, 128 experts), we isolate three evaluation axes that change conclusions, not just numbers. Replay semantics: under a fused-event traffic contract, an inconsistent per-access replay inflates recency-based policies by 27-29% while leaving frequency-based and static ones within 4%, inverting the policy ranking. Workload contamination: probe sets using one instruction template per category produce verbatim-identical generation prefixes; a matched-pair rendering intervention moves the measured early-window effect by 19.4-31.9 points and reverses which workloads look most cache-friendly. Operating regimes: normalized miss fractions do not transfer across models, so the per-step expert union relative to per-layer capacity must be reported — yet permuting only the temporal order of an identical event stream moves the offline-optimal gap from 44.9% to 30.8%, so it is not sufficient. Corrected, a stable gap to the offline optimum remains (44.2-45.9% over 13 frozen workload compositions). A forced-admission oracle attributes 84.3-96.6% of it to knowing which resident expert is used furthest in the future. A causal next-use predictor, used as an eviction rule, recovers -11.4% of the gap; it picks an optimal victim 3.4% of the time, against 2.4% for a random resident block and 20.6-22.1% for LRU and LFRU. Our position is narrow: in our evaluated settings a large offline-optimal gap substantially overstates the gains recovered by representative lightweight causal mechanisms.

Agentic retrieval-augmented generation (RAG) systems combine preprocessing, embedding, retrieval, memory access, context construction, generation, and vector-index updates. Although LLM decoding is GPU-bound, the surrounding orchestration layer can still limit end-to-end performance through serialization overhead, fragmented scheduling, inefficient batching, and CPU--GPU pipeline stalls. Existing frameworks provide flexible control flow, while distributed runtimes provide scalable task parallelism, but neither exposes RAG stages as resource-aware operators with deterministic execution semantics. We present OpRAG, a resource-deterministic distributed runtime for GPU-backed multi-stage RAG workflows. OpRAG models embedding, retrieval, reasoning, memory, and upsert as first-class operators and lowers them into communication-aware execution graphs. It combines an Arrow zero-copy data plane, persistent workers, bounded queues, CPU tokenizer prefetching, batched GPU embedding, and overlapped retrieval/generation execution to reduce non-model overhead around LLM inference. We evaluate OpRAG using Llama3-8B and Mistral-7B with FlashAttention~2, BF16 execution, and 32K RAG chunks. In end-to-end GPU pipeline experiments, OpRAG improves over the nearest competitor by 16.16% for Llama3-8B and 15.66% for Mistral-7B, and over RayScalableRAG by 20.57% and 20.71%, respectively. Against LangChain, LangGraph, CrewAI, and AutoGen, OpRAG is 17.77% and 17.48% faster than the best framework baseline. In Higress-style query serving, OpRAG reduces hybrid retrieval latency by 59.20--59.62% and generation-scenario latency by 52.48--53.55%, while preserving 100% Recall@5. These results show that optimizing the distributed orchestration layer can substantially improve GPU-backed multi-stage RAG without modifying the LLM decoding kernel.

Large language models can answer knowledge-intensive questions more reliably when they are grounded with knowledge graphs, but systems such as Think-on-Graph and Reasoning-on-Graph repeatedly query the same graph neighborhoods across different questions. In this work, we study this repeated retrieval in Knowledge Graph Question Answering~(KGQA) workloads and propose KGCache, an in-memory cache for one-hop knowledge graph neighborhoods. KGCache is designed to be compatible with both iterative traversal (ToG) and one shot planning (RoG) KGQA paradigms. KGCache is placed between the KGQA engine and the backend serving the KG, so repeated entity requests can be served from cache instead of issuing new KG queries. We evaluate KGCache on WebQSP and CWQ using LRU, LFU, and a trace-aware Oracle policy. Our analysis shows that both datasets contain substantial entity reuse among starting entities and entities reached during traversal. We also explore semantic caching for similar queries, which shows additional hit-rate gains on WebQSP and needs further accuracy testing on CWQ. Entity caching accelerates KG retrieval by up to $1.91\times$, while semantic-context caching achieves up to $1.06\times$ full-system speedup in the evaluated WebQSP configurations, with each hit being up to $3.73\times$ faster.

Multi-turn Reasoning-and-Acting (ReAct) agents accumulate growing trajectories of reasoning, tool calls, and observations. Their key-value (KV) caches grow accordingly, increasing memory use and attention cost during model inference. Existing KV cache compression methods reduce these costs by evicting states with low attention scores. However, low attention in the current turn does not imply future irrelevance, as temporarily inactive information may become important later. Snapshot-based eviction methods therefore do not explicitly distinguish temporarily dormant information from information that appears to have completed its role. In this paper, we present CommitKV, which identifies KV lifecycles through commit transitions. Specifically, CommitKV first divides completed agent events into token pages and compares each eligible page's deletion effect before a tool-call commit and after the commit's returned observation has been incorporated. Based on these paired measurements, CommitKV distinguishes dormant pages from high-to-low completion candidates. It then applies a greedy joint test, accepting candidates for retirement only when their combined post-commit effect remains bounded. Finally, at a later compression checkpoint, accepted pages are excluded, a bounded set of pages awaiting post-commit measurement is protected, and the remaining KV states are retained within the cache budget using the same token indices for keys, values, and absolute positions. These mechanisms ensure that CommitKV can distinguish dormant information from information that has completed its observed role and can be safely removed. Experiments on various benchmarks show that CommitKV reduces agent memory use, accelerates end-to-end inference, and achieves higher accuracy than existing KV cache compression methods.

Large mixture-of-experts (MoE) language models with 26--120 billion parameters exceed the memory capacity of consumer devices through three simultaneous pressures: resident weight matrices, key-value (KV) cache state that grows linearly with context, and dozens of expert sublayers that must be paged on demand. We present RotaryQuant, a three-axis compression system that addresses all three. Mixed-precision weight quantization assigns bit-widths by architectural role: 4-bit for dense layers, 2-bit for routed experts, and 8-bit for the shared expert whose high activation kurtosis resists aggressive compression. LRU expert offloading pages non-resident experts to disk under genuine memory pressure. The novel axis is IsoQuant, a KV cache compression method that applies a Walsh--Hadamard transform followed by block-diagonal SO(4) rotations to isotropize activation distributions before 3-bit scalar quantization, requiring $O(d \log d)$ operations and 256 stored parameters per head versus $O(d^2)$ and 16{,}384 for dense rotation methods. A fused four-kernel Metal GPU pipeline performs attention directly on packed 3-bit tensors without materializing full-precision KV state---a different execution model, not just a quantization scheme. The combined system fits Gemma 4-26B-A4B and Qwen3-30B-A3B within a 16\,GB budget and Nemotron-H 120B within 32\,GB, running interactively at 9--19 tok/s with near-zero perplexity degradation ($Δ$PPL $\leq +0.0012$) and 100% retrieval accuracy at 32K context.

Large language models (LLMs) increasingly read long inputs in the agentic era, from whole documents and codebases to conversations across many turns. Their inference memory is then dominated by the key-value (KV) cache, the stored attention keys and values of every token the model has read and generated. Because the cache grows with context length and is re-read in full at every generated token, a longer context means more GPU memory. To reduce this cost, most existing methods compress the KV cache by lowering every stored value to the same low precision, a technique known as quantization. They can push this to nearly two bits per value, but rarely further, because quality drops sharply at this 2-bit cliff: four levels are too few for the cache's outlier-heavy values, where a few large entries consume the levels and collapse the rest into noise. A natural remedy is to spend more bits on the channels (feature dimensions) that matter and fewer on the rest, but the raw cache offers no handle: its channels are strongly correlated, so none stands out as more important. Our analysis shows that this handle appears once the cache is rotated into a coordinate system computed from its own statistics, removing these correlations. There, a small fraction of channels carries almost all the information, and spending the budget on those few is far more accurate than spreading it evenly. Guided by this analysis, we develop SPECTRA, a training-free, drop-in codec that re-encodes the cache into this coordinate system and concentrates the bit budget on the channels that carry the signal. On Llama-3.1-8B and Qwen2.5-7B over long-context benchmarks, SPECTRA is near-lossless at 4x compression, competitive at 8x where uniform quantization has collapsed, and reaches up to 12x, pushing usable compression past the 2-bit cliff so the same GPU holds longer contexts and larger batches.

Large language model (LLM) inference serving is increasingly constrained by memory rather than compute. As long-context and long-form reasoning workloads become more prevalent, the key-value (KV) cache dominates both memory footprint and memory traffic during LLM token generation, i.e., decode. In particular, HBM capacity has become a scarce and costly resource that heavily limits inference batch size and system throughput. This paper presents OasisKV, a memory-centric LLM inference system design that alleviates HBM capacity pressure by decoupling full KV-cache storage from HBM during LLM decoding. Because decode-time attention is naturally sparse, OasisKV keeps only the KV entries of the most relevant tokens in HBMs for attention computation. We observe that future important tokens can be predicted accurately in advance using lookahead tokens drafted by speculative decoding (SD). OasisKV employs an efficient attention background pipeline to identify important KV blocks. They are then prefetched from higher-capacity memory tiers (e.g., host or remote memory) and staged in HBMs before being used in the next decode step. We implement OasisKV based on vLLM. The lookahead prediction is accurate enough to keep accuracy within 0.7 points of full attention under a 2,048-token KV budget. This lets OasisKV turn sparsity into throughput gain: $1.69\times$ over dense vLLM on the reasoning workload at 0.1 points of accuracy loss, and up to $2.1\times$ on multi-GPU long-context serving. Under prefill--decode disaggregation, OasisKV reaches about $2\times$ dense throughput while admitting each request with $6.5$--$9.7\times$ less KV and holding $2.2$-$2.6$ less decode-node host memory than full KV transfer.

Diffusion language models (DLMs) iteratively refine a sequence, allowing earlier predictions to be revised as context evolves. This rollback capability distinguishes them from irreversible autoregressive generation, but makes inference costly. Every denoising update alters the global context, forcing both prompt and response states to be recomputed even though only response tokens are revisable. Key-value (KV) caching could reduce this cost, yet conventional caching assumes immutable historical states and is therefore difficult to reconcile with rollback. In this paper, we introduce Adaptive Reuse of Cached Hidden States for Efficient Rollback (Archer), a training-free KV caching method for rollback-capable DLMs. Archer asymmetrically keeps the mutable response synchronized with the current hypothesis while reusing prompt K/V within a bounded state neighborhood. Although prompt representations also change under bidirectional attention, their token identities remain fixed; bounded reuse therefore amortizes repeated prompt computation without caching mutable response states. It also delays feedback from tentative tokens, reducing premature reinforcement of transient high-confidence errors and giving rollback more opportunity to correct them. Our analysis characterizes prompt reuse as a reversibility-aligned cache boundary, bounds its state-dependent approximation error, and gives a decoder-margin condition for preserving full-refresh decisions. Existing DLM acceleration often trades quality for speed. Archer shifts this frontier, attaining the best mean performance of 33.63% together with a 2.57x mean speedup on the main suite. Across evaluated settings, it improves Pass@1 by up to 3.05 points and reaches up to 2.95x speedup. Controlled analyses connect the quality gain to delayed prompt feedback and validate state-aware refresh. Our code is available at https://github.com/Hxnng/Archer.