PyTorchKR Papers

최신 논문

총 523,801편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2611.04337v1cs.CV

Mitigating Spatial Blindness in Multiple Instance Learning for Whole-Slide Image Analysis

Whole-slide multiple instance learning can predict diagnostic and prognostic labels from tissue composition while making limited use of spatial organization. Providing coordinates or a graph does not establish that a trained predictor uses tissue neighborhoods. We examine this gap as spatial blindness and hypothesize that early composition learning can weaken supervision of a spatial pathway. We introduce ResTopoMIL, which freezes a learned composition predictor before training a graph residual with a coordinate-reassignment separation loss. Permutation and information analyses distinguish architectural symmetry, available spatial information and the proposed optimization mechanism. On a controlled spatial task, staged training achieved an AUC of 0.987, compared with 0.684 for the same jointly trained architecture. Performance remained high on a composition-only task. Across four WSI classification and five survival benchmarks, ResTopoMIL improved mean primary performance on eight tasks; four improvements remained significant after Holm-adjusted patient-cluster tests. Coordinate-free residual controls recovered some of the gains, while neighborhood destruction reduced performance close to the composition anchor. Neighborhood-preserving transformations and anatomy-stratified perturbations further characterized this dependence. These results support fixing a composition reference as a practical intervention for learning spatial corrections, while retaining additional appearance capacity as a contributing explanation.

arXiv:2611.04330v1cs.AI

CORA: Teacher-Guided Causal O-RAN Adaptation for Jamming Detection

In mission-critical networks, a missed jamming event is costlier than a false alarm, so detectors should favor recall, operate causally, and work without labeled attack data. Retrospective Key Performance Indicator (KPI) detectors meet the label-free requirement but rely on future context and cannot run in streaming operation. We present CORA, a teacher-guided reinforcement learning (RL) framework that transfers the hindsight of a retrospective detector to a causal one without labels during training or deployment. Offline, a retrospective KPI detector evaluated on unlabeled historical logs defines a recall-weighted reward for a masked controller that adapts the sensitivity, persistence, and grace of a causal KPI detector. At deployment, the teacher is removed and the frozen controller acts causally, with no ground-truth labels and no target-specific calibration. The design maps onto offline training in the Non-Real-Time RIC and causal control in the Near-Real-Time RIC of an Open Radio Access Network (O-RAN). Under leave-one-out transfer across ten 5G O-RAN testbed configurations, CORA reaches recall-oriented operating points on the deployable frontier that none of the evaluated baselines covers. A Proximal Policy Optimization (PPO) controller raises mean recall from 0.687 to 0.771 over a fixed default detector on all ten configurations, while a Deep Q-Network (DQN) achieves a comparable F2-score to the best teacher-selected static configuration, with 61% fewer false alarms. Removing the teacher lowers recall on every configuration.

arXiv:2611.04327v1cs.LG

Anchor: Augmented Cross-Layer Recurrent State for Long-Context Retrieval in Test-Time Training

Parameter-efficient linearization turns a pretrained autoregressive Transformer into an efficient model by hybridizing sliding-window attention (SWA) with a constant-size recurrent memory. But does every layer's recurrent state contribute to long-context retrieval? Investigating an SWA and Test-Time Training (TTT) hybrid, we find that retrieval is sparse across depth, concentrated in a few anchor layers, and that the remaining states hold similar information expressed through layer-specific transformations. We exploit both observations in Anchor: TTT memory placed only at the anchor layers, and a cross-layer recurrent state written once and read by many layers, each through its own learned read map that realizes a layer-specific stretch and rotation, along with demonstrated generalization to other hybrid attentions. On Llama-3.1-8B and Mistral-7B, Anchor delivers 2.3-3.1x higher recall and 6.7-8.9 points higher MMLU accuracy than LoLCATs with a 10x smaller recurrent state, retains 91-95% of the teacher's MMLU accuracy, and cuts end-to-end latency by up to 4.3x relative to softmax attention.

arXiv:2611.04323v1cs.AI

FP4Quant: Efficient 4-Bit Communication Compression for Model-Parallel LLM Training

The rapid scaling of large language models (LLMs) has made communication overhead a critical bottleneck in distributed model-parallel training. Although existing low-precision compression methods can effectively reduce communication costs, further reducing the communication footprint to ultra-low-bit formats such as FP4 introduces new challenges, including limited numerical range, sub-byte storage overhead, and increased computational cost from fine-grained quantization. In this paper, we present \textbf{FP4Quant}, a micro-grouped FP4 compression framework for efficient model-parallel LLM training. FP4Quant integrates a micro-grouped dual-scale FP4 quantization mechanism to preserve local tensor distribution characteristics, a warp-cooperative FP4 packing scheme to enable efficient sub-byte data representation, and a hybrid two-level FWHT pipeline to mitigate quantization errors with GPU-aware execution optimization. Experiments on NVIDIA H100 GPUs with Qwen-7B demonstrate up to 1.56$\times$ end-to-end throughput improvement over BF16, revealing a trade-off between throughput and model quality.

arXiv:2611.04321v1cs.AI

ATC: Agent Trajectory Control for Role-Contract Enforcement in Black-Box Multi-Agent Systems

Multi-agent LLM systems assign different responsibilities to different agents: one plans, another retrieves evidence, another verifies, another executes. This division of labor is only meaningful if each agent stays within the functions it is authorized to perform. Over many turns, however, an agent can produce a locally reasonable message that exercises authority belonging to another role: a verifier that begins planning, a retriever that asserts unsupported conclusions, an executor that changes state no one approved. We formalize this as role drift, a turn-level breach of an agent's role contract. We introduce Agent Trajectory Control (ATC), which makes each role's authority explicit as a contract and enforces it at the point where a message would enter the shared conversation. ATC scores a proposed output for out-of-role deviation, leakage into the functions of other roles, impermissible speech acts, and logical contract breaches, then computes a leakage-regularized target that constrains regeneration; the revised message is re-checked before admission. Enforcement acts on the message rather than the generator, so ATC governs agents whose internal representations are unavailable. Across homogeneous and heterogeneous teams, ATC improves task performance and role consistency while lowering role-drift energy relative to strong baselines spanning prompting, memory, workflow control, and activation editing, with the largest gains under explicit role-confusion stress tests and in larger, more diverse teams.

arXiv:2611.04320v1cs.AI

Planning-Coding Separation in LLM-Based Code Generation: An Empirical Study

Planning has been identified as an important factor for improved code generation with Large Language Models (LLMs). However, the interaction between planning capability and coding capability remains largely underexplored, especially when the two capabilities are from different LLMs. In this paper, we conduct the first comprehensive empirical study of planning-coding interaction in LLM-based code generation, where planning and coding are assigned to different LLMs (i.e., planning-coding separation). Specifically, we explore different planner-coder combinations to investigate the effectiveness for coders to utilize planning information from various planners and the impact of further training of coders on planner-coder combinations. Our results demonstrate that almost all planner-coder combinations are as competitive as using the same LLM for planning and coding, and the combination of a stronger planner and a stronger coder typically achieves improved effectiveness. Furthermore, using a large model as the planner paired with a much smaller model as the coder can sometimes be as competitive as using the large model as both the planner and the coder, and further training of the coder can typically boost various planner-coder combinations. These findings provide empirical evidence that planning and coding can be separated to a considerable extent in LLM-based code generation, which may offer a more flexible way to construct code-generation systems.

arXiv:2611.04318v1cs.DB

Text-to-SQL Models Make Choices Without Saying So

A text-to-SQL answer can run, return a plausible number and still compute something other than what the analyst meant. This can happen when the request leaves a calculation choice open, such as whether to count rows or distinct entities, and the model settles it without saying so. Standard benchmarks cannot detect this, because they score each question against a single reference interpretation. We introduce SpecBridge, which tests such a choice by varying it alone: it poses three otherwise identical requests, one leaving the choice open and one per interpretation, and scores each answer by running it. The databases are real, and the choices are ones that made models fail on existing benchmarks. On 204 hand-validated tests, four language models usually compute each interpretation when asked for it, yet for the open request return both in only 15-58% of the tests where they can, often without mentioning the choice. The gap lies in reporting, not in computing. Benchmarks even penalize the alternative: on BIRD questions where such a choice changes the answer, over two in five of the answers marked wrong are right under the other interpretation. Because the gap lies in reporting, a system can narrow it without a stronger model: it first decides whether a request leaves a choice open, then answers each interpretation in its own call. On our tests and on AMBROSIA, an existing benchmark of ambiguous questions, this approach raises the share of open requests that get every interpretation by 29-63 points over the single answer that production assistants give, without lowering precision on clear requests. It also needs fewer user turns than asking a clarifying question.

arXiv:2611.04317v1cs.CV

SteelHuman: Predicting Operator Estimates of Steel-Scrap Composition from Sparse RGB Views

Steel scrap composition estimation is an important part of yield planning and fair supplier compensation. Measuring the composition of a multi-ton pile is an inherently laborious process, so the assessment is often left to trained human operators. Recent works tackle composition-level tasks, but require structured acquisition and multi-stage preprocessing. We study a simpler approach to steel scrap load assessment by aggregating features from sparse, handheld views of the pile. On 600 out-of-fold piles, frozen DINOv2 tokens aggregated uniformly across views reach 13.4% total variation (TV) distance against operator estimates. Most remaining error lies in allocation among four thickness-defined heavy melting scrap (HMS) classes, yet image-dependent allocation reduces within-HMS TV from 25.1% under a training-set split prior to 11.2%. Due to the lack of public datasets and benchmarks for predicting composition, external evaluation relies on the public DOES dataset intended for classification. A classifier using the same frozen features reaches 93.97% accuracy, exceeding the published results we compare against. The composition model also aligns with related DOES categories without adaptation.

arXiv:2611.04315v1cs.AI

Drift No More: Spectral Locking for Long-Horizon Open-Weight Multi-Agent AI

Long-horizon multi-agent systems suffer from three tightly coupled failures: task drift, role drift, and state staleness. We test the hypothesis that these share a common representational cause: agent hidden states move away from the directions encoding the active task, assigned role, and current shared state. A pilot study over 14,832 annotated turns supports this: drifted turns collapse near-uniformly across all three anchor subspaces (pairwise r = 0.72-0.78). Existing methods address these failures separately, operating at the surface-text level or treating control directions independently. We introduce ATLAS (Algebraic Task-role Locking with Adaptive Spectral Memory), a training-free intervention acting on hidden states before a token is generated. ATLAS builds a positive-semidefinite anchor kernel over task, role, and state subspaces and steers hidden states toward their shared high-coherence eigendirections; a low-rank Gram decomposition keeps this tractable. Across long-horizon memory, state maintenance, tool use, least-privilege coordination, and embodied stress tests, ATLAS outperforms prompting, retrieval, memory, and activation-steering baselines. Against A-MEM it improves memory macro-average accuracy by up to 5.7 points and reduces write/read failures by up to 31.7%/24.8%; against MemoryArena-Agent it gains up to 4.8 points. It gains 10.7-12.1 points under perturbation, reaches 93.5% role consistency, and transfers to a 72B backbone (macro-average 61.0 vs. 56.2). Ablating the joint task-role-state span costs 7.2 points against 0.8 for a matched-rank random span, indicating these directions are functionally implicated, not incidental; supplying them by prompt or retrieval recovers 4.3 points less. ATLAS is therefore a representation-level controller for open-weight multi-agent systems, complementary to output-level protocols for black-box agents.

arXiv:2611.04313v1cs.LG

Coarse Semantic Blocks for Efficient Training-Free Sparse Attention

We introduce a training-free sparse attention method that builds a prompt-specific key index to accelerate both prefill and decoding. The method clusters the keys into large semantic blocks that group tokens from noncontiguous positions. Each query position selects blocks by scoring their centroids and attends to the original keys and values within them. Queries that select the same block are processed together, so prefill stays fast even though each query chooses its own blocks. Large blocks make centroid scoring and this grouped processing cheap, so each query can attend to more tokens in the same time. Larger blocks with a larger token budget can thus be both faster and more accurate on long-context tasks than smaller blocks with a smaller budget. We evaluate two hybrid LLMs, Qwen3.5-4B and 35B-A3B, on one NVIDIA H200. Including the cost of building and using the index, whole-model wall-clock speedups over dense attention are 2.3-2.4x for prefill and 1.7-2.0x for decoding at 512K tokens, and 2.9-3.0x for prefill at 1M. The 35B-A3B model stays close to dense attention on five RULER tasks through 1M, on LongBench-v2 overall, and on MRCR through 256K. The 4B model loses accuracy mainly on multi-key retrieval, yet outperforms dense Qwen3.5-2B, whose latency is similar at 512K, on all three benchmarks.

arXiv:2611.04309v1cs.LG

What Is the Influence of Task Order in Continual Learning? An Empirical Study and Order-Aware Evaluation for Motion Forecasting

Aggregate scores for continual learning algorithms rest on a distribution over task orders that is rarely stated. We formulate algorithm comparison as an expectation over a declared order distribution, which separates training variation from order variation, and apply it to motion forecasting with four road-layout tasks from the INTERACTION dataset, all 24 task orders, four algorithms, and three training replicates (288 runs). Task order accounts for 65\% to 91\% of the variation in current-task error and backward transfer, and 90\% to 99\% of this order effect comes from the identity of the starting task: the smallest task cannot be learned from a cold start. Beyond the first stage, the order effect is comparable to training variability. In most orders, the leading algorithm in current-task error and backward transfer is within one replicate standard deviation of the runner-up. Backward transfer depends on the order mainly through its reference term, which weights the first task most heavily, so relative improvement and final error identify different leading algorithms for displacement error. A transition-direction descriptor computed from task statistics tracks the position of the smallest task, and weighting orders by it leaves the algorithm ranking unchanged. These results call for benchmarks that declare their order distribution and report order-resolved, starting-task-resolved, warm-start, and final errors alongside aggregate scores. The code and data are available \href{https://anonymous.4open.science/r/task-order-cl-4f7b/}{here}.

arXiv:2611.04307v1stat.ML

Nearly Linear-Time Follow-the-Perturbed-Leader for Weakly Observable Graph Bandits

We study online learning with directed weakly observable feedback graphs and propose a Follow-the-Perturbed-Leader (FTPL) policy that achieves best-of-both-worlds guarantee with nearly linear expected time complexity. To preserve the simple structure of FTPL, we develop a novel graph-based loss estimator that combines forced exploration, conditional geometric resampling and adaptive learning rates with surrogate probabilities. Our policy achieves $\tilde{\mathcal{O}}(\delta_*^{1/3} T^{2/3})$ regret for adversarial regimes as well as $\tilde{\mathcal{O}}(\delta_* \log T / \Delta_{\min}^2)$ regret for stochastic regimes, where $\delta_*$ and $\Delta$, respectively, denote the fractional weak domination number and the suboptimality gap. These results show that refined guarantees for problems with minimax regret of $\Theta(T^{2/3})$ can be obtained through appropriately designed adaptive learning rates and loss estimators without sacrificing the simplicity of FTPL.

arXiv:2611.04305v1cs.CV

Weather Adaptive Autonomous Driving

Camera object detectors lose recall in rain, fog, snow and sand, and the confidence threshold at which a detector is deployed decides how many objects it misses and how many false alarms it raises. We study how a frozen detector should make this decision in adverse weather. We first reproduce SeqCRC, a conformal object detector whose first step calibrates the confidence threshold by conformal risk control (CRC), and use that step, calibrated separately in each weather condition, as our baseline. With YOLOv8x it keeps the held-out miss fraction within a 20% budget in every DAWN weather bin, where a single global threshold fails in fog, but the false alarms it needs are an output rather than a choice: from 1.6 to 117 per image across DAWN and BDD100K conditions. We then propose a calibrated trade-off dial. One setting, $\lambda$, the number of false alarms accepted to avoid one missed object, selects the expected-cost-optimal decision rule: either one score threshold, or a per-detection rule that keeps a box when its calibrated probability of being a real object, estimated by gradient-boosted trees from the detector's own per-box outputs, is at least $1/(1+\lambda)$. F-score preferences map onto the same dial, and CRC applied to $\lambda$ itself is planned to restore a per-condition miss guarantee. The dial is fitted once on BDD100K train and graded only on BDD100K val and DAWN.

arXiv:2611.04303v1cs.CV

FlowTranslator: Training-Free In-Image Machine Translation with Flow-Based Editing

In-image machine translation (IIMT) replaces text in an image with its translation while preserving the surrounding visual context. Existing task-specific models achieve convincing performance but require massive paired training images before and after translation. As advanced text-to-image (T2I) models have reached a practically usable level of multilingual text generation, we argue that this capability can be transferred to IIMT and present FlowTranslator. It is a training-free IIMT framework that first employs an OCR model and an LLM to generate the translation prompt, which is injected into the T2I-based diffusion model. To optimize the accurate and natural generation of translated text, we introduce two trajectory controls: Region Coupling Release (RCR) partially decorrelates target-branch noise inside the text region early in the trajectory, while Coupling Expansion (CE) amplifies the emerging edit displacement mid-trajectory. Under the EasyOCR read-back protocol, FlowTranslator achieves average PRIM BLEU and COMET scores of 20.6 and 57.2, respectively, substantially outperforming the strong training-based baseline UniTranslator. On the more challenging and realistic whole-image ScenePair photographs translated into Chinese, it reaches 40.1\% exact-match accuracy and surpasses existing editing systems, some of which only address the text-rendering stage rather than the complete IIMT task. These results establish flow-based editing as a strong, fully training-free baseline for IIMT.

arXiv:2611.04301v1cs.AI

OSGR: Observation-Support Graph Reasoning for Earth Observation Question Answering

Answering geographic queries over Earth observation (EO) archives requires solving a joint retrieval-and-derivation problem under strict spatiotemporal constraints. Existing LLM agents fail to ground observation items or suffer from search-space explosion. We propose OSGR, a two-stage framework that factorizes this search into offline symbolic concept deduction and online Relation-Conditioned Grounding. OSGR leverages an offline Observation-Support Graph to deduce valid target-to-result support paths at the concept level, and applies spatiotemporal compatibility kernels to instantiate grounded observation-item sets. On Earth-Bench and DisasterM3-Reason, OSGR improves GPT-4o accuracy from 49.3\% to 63.7\% while reducing the execution-step ratio from 2.9 to 2.5.

arXiv:2611.04300v1cs.CV

CLOCs-R: Radar-Assisted Camera-LiDAR Fusion for Long-Range Object Detection

Long-range 3D object detection remains difficult for autonomous vehicles because LiDAR returns become sparse and camera evidence weakens with distance, while automotive radar keeps returning measurements far beyond both. We present CLOCs-R, an extension of the CLOCs candidate-level fusion framework that adds radar evidence to camera-LiDAR rescoring. Radar returns are accumulated over several sweeps, compensated for ego-motion, and associated with each pre-NMS 3D candidate through a range-dependent gate. Compact features summarizing support, Doppler speed and spread, range residual, and offset are then appended to every camera-LiDAR pair before a lightweight fusion network rescores the candidates. We first verify that our fusion implementation reproduces CLOCs on KITTI. We then evaluate on the daytime clear-weather subset of the aiMotive dataset, reporting distance-binned metrics out to 200 m with frozen SECOND and Cascade R-CNN detectors. Camera-LiDAR CLOCs improves on SECOND up to 50 m but degrades beyond it. A residual scoring variant, introduced post hoc, corrects SECOND's score instead of replacing it and recovers most of this loss. Radar consistently improves BEV AP at 30-50 m (0.857 to 0.870), but beyond 50 m its gain is small (+0.003 center-distance AP), within seed variation, and CLOCs-R with residual scoring is statistically tied with LiDAR-only SECOND. Our analysis attributes the limited long-range benefit to weak separation of true and false detections by radar features, scarce far-range training positives, poor camera evidence at distance, and missing LiDAR proposals that rescoring cannot recover. We report all as-designed and post hoc results side by side.

arXiv:2611.04296v1cs.HC

SEAM: Situated Executable Associative Memory for a Co-located Community

In groups, knowledge is distributed among individuals, and how it is brought together depends on social ties. Taking inspiration from how a single mind recalls and reacts to memories and current stimulus, we build a shared memory that does this for a community, prototyped as a social app, SEAM. Members post arbitrary moments; SEAM works out how fragments relate, distills knowledge about the group, and replies in a situated and associative way. Some of these memories act as instructions that can alter how the system behaves, making it programmable by its members. We deployed SEAM to 46 people in a co-located community for two weeks of open-ended use, and ran a follow-up workshop with 14. Members shared their days, interrogated the memory, and instructed it to track routines and run games; replies introduced them to people and subjects they had never encountered. SEAM points towards communities where knowledge finds people.

arXiv:2611.04295v1cs.LG

Reciprocity Can Erase Direction:tructural Boundaries and Controlled Utility of Magnetic Graph Networks

Magnetic graph networks encode direction through phase, but a learned graph can erase the intended directional mechanism before prediction. We characterize this failure for support-based magnetic Laplacians. Exact identities relate imaginary energy and distance from zero phase to amplitude-weighted reciprocity. Top-$k$ selection forces complete collapse when the observed-node count satisfies $m\leq k+1$, and dense support imposes a lower bound on reciprocal-arc fraction. We then construct paired propagation examples with identical node features and symmetric amplitudes but opposite labels, giving a provable chance-level limit for direction-erasing representations. In 72 fixed neural fits, support phase achieved test AUROC 1.0000 on a unilateral ring but 0.5000 on reciprocal ring and complete supports; continuous score phase and a real directed comparator achieved 1.0000 throughout. All models solved the direction-free control. These are oracle-direction mechanism tests, not evidence of learned clinical direction. In the preserved public DeepDRiD development study, continuous phase was active but its robustness difference from zero phase was only $+0.0036$ (descriptive 95\% interval $-0.0014$ to $+0.0095$) on validation patients used for checkpoint selection. The contribution is a structural and experimental framework for distinguishing erased direction from unestablished task utility. It provides a precise circumstance in which magnetic phase preserves useful information, without asserting universal or clinical superiority.

arXiv:2611.04292v1stat.ML

How Risky Is Your Data-Driven Policy?

Data-driven algorithms are ubiquitous across industries, and there is an extensive literature on analyzing their \emph{average} performance. The risk profile of such algorithms, while critical, is much less understood. This paper quantifies the risk of data-driven policies and isolates its sources. We evaluate a policy through the conditional value-at-risk (CVaR) of its loss over the training sample, any policy randomization, and the future outcome, at a fixed unknown distribution. Regret against an oracle that knows the distribution decomposes into expected post-sample regret (PS), the average outcome-only CVaR gap of the selected action, and a sample-risk premium (SR), the additional tail risk omitted by averaging those evaluations. The analogous premium vanishes under expected loss. We characterize when the premium vanishes and how large it is relative to post-sample regret. For losses with a continuous density near the tail cutoff, we identify conditions under which both components decay at the same rate, yet the limiting ratio of the premium to post-sample regret can reach the thousands. When a probability mass at the cutoff moves with the action, the premium can decay more slowly and dominate post-sample regret. We establish conditions under which sample average approximation learns the oracle action at rate $n^{-1/2}$ in both settings, but its combined-risk regret is of order $n^{-1}$ in the first and $n^{-1/2}$ in the second. In finite-sample numerical examples of SAA in both settings, the premium accounts for approximately $74\%$--$97\%$ of regret.

arXiv:2611.04289v1cs.CR

When Debate Amplifies Risk: Social-Bias Vulnerabilities in Multi-Agent Safety Deliberation

Retrieval-augmented Web applications increasingly employ multi-agent debate (MAD) to aggregate conflicting, noisy, and heterogeneous information, yet deliberation is fundamentally a social process susceptible to interpersonal dynamics. When retrieved Web sources contain latent hazards alongside safe factual content, social biases such as conformity pressure and authority deference can override evidence-grounded safety boundaries. In this work, we investigate whether the social structure of deliberation can paradoxically amplify latent Web safety risks. We introduce SafeConflict-Bias, a controlled counterfactual diagnostic protocol spanning six core Web risk categories, which holds queries and retrieved evidence strictly invariant while isolating conversational majority sway, authority cues, and confidence asymmetry. Empirical evaluation across diverse proprietary and open-source backbones reveals that while isolated models reliably reject unsafe guidance, standard MAD catastrophically capitulates to social pressure---surging unsafe adoption from 0.0\% up to 100.0\% and systematically silencing safe minority evidence. To resolve this systemic fragility without retraining or debate regeneration, we propose RiskTrace-Debias, an inference-time auditing framework uniting evidence-first precommitment, social-metadata masking, and consistency-gated arbitration. RiskTrace-Debias suppresses worst-case unsafe adoption to $\le 2.0\%$ across all architectures without inducing over-refusal, while generating verifiable provenance traces to audit interpersonal manipulation in Web deliberation.