PyTorchKR Papers

최신 논문

총 523,801편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2611.04886v1cs.AI

Modelling lock-in of AI adoption through skill erosion

Generative AI can improve immediate performance but may erode the human skills required for future work. We study a model in which individuals tend to imitate successful peers, while their independent skills grow with practice and erode with reliance on AI. Because each agent's skill reflects their history, non-users and AI users develop different skill distributions, which we derive exactly. Skill erosion can produce bistability and hysteresis between a low-adoption, high-skill state and a high-adoption, low-skill state. When agents rarely change behaviour at random, the two transitions are asymmetric. A single AI user spreads through a population of experts only if AI-assisted work earns more than expert work, whatever the timescale of skill adaptation. A single non-user, by contrast, can lead an AI-dependent population back to independent work only if they regain enough skill before returning to AI, which becomes easier the more persistent behaviours are. As AI improves, adoption can jump abruptly, and a sufficiently long increase in its benefit, such as a period of low prices, can leave a population relying on AI after the benefit has returned to its initial value. Skill erosion thus creates a history-dependent switching cost that can make widespread AI adoption difficult to reverse.

arXiv:2611.04885v1cs.LG

DLER: Dual-Logit Experience Replay for Online Continual Learning

Online continual learning aims to continuously train neural networks on a non-stationary data stream in a single pass. Rehearsal-based methods alleviate catastrophic forgetting by replaying stored samples, but the imbalance between current and previous data often leads to predictions biased toward current classes. Many approaches have been proposed to overcome the forgetting problem by correcting the bias; however, they still need to be improved in an online fashion. By decomposing the dot-product logits into an angle factor and a norm factor, our analysis reveals that the angle factor efficiently captures novel knowledge but is particularly susceptible to prediction bias, whereas the norm factor helps preserve historical knowledge. Based on these observations, we propose Dual-Logit Experience Replay (DLER), which exploits dual-logits for both model optimization and memory management. Specifically, DLER learns current samples using cosine logits and incorporates additional dot-product logits when replaying previous samples to mitigate prediction bias. To strengthen dual-logit learning under limited replay memory, an auxiliary supervised contrastive objective promotes angular compactness within each class, extending corrections learned from limited replay samples to other samples of the same class. Meanwhile, we introduce a memory update strategy that preferentially retains samples for which dot-product prediction assigns a higher probability to the correct class than cosine prediction, aligning sample selection with dual-logit training. Extensive experiments on three datasets demonstrate the effectiveness of DLER, particularly under limited memory budgets.

arXiv:2611.04874v1cs.CV

OCR in the Foundation Model Era: From Specialist Models to Unified Models

Optical Character Recognition (OCR) is among the oldest and most widely deployed applications of artificial intelligence. This survey reviews its development, focusing on the Foundation Model Era. We define OCR through faithful transcription: within a specified visual scope, OCR should completely reconstruct the text, notation, and document structure observable in the source image, and every textual, notational, or structural element in the output must be verifiable from visual evidence in that image. We trace the development from specialist OCR models toward unified models, examining joint text detection and recognition, structured sequence generation, and the use of pretrained vision-language models (VLMs). We organize contemporary OCR systems into five coexisting families: modular toolkits and task-specific models, specialized document VLMs, hybrid workflows, general-purpose VLMs, and commercial products and managed services. We compare their architectures, data, and training, examining hallucination and omission, insufficient spatial grounding, and inference cost in generative OCR. We review evaluation from character and scene text recognition to parsing individual structural elements and full-page transcription, highlighting gaps in consistent output definitions, attribution of performance to individual components, and coverage of real capture conditions and visible content and structure. Finally, we summarize the challenges facing OCR and discuss directions for future research. We continuously track related work at https://github.com/zhihengli-casia/Awesome-OCR-in-the-Foundation-Model-Era.

arXiv:2611.04873v1cs.RO

Trusting Imagined Futures: Auditing Imagined Rewards for VLA Post-Training

Video world models enable candidate actions to be evaluated through imagined outcomes, but high predicted rewards can favor trajectories with poor execution quality. We introduce Trusting Imagined Futures (TIF), an offline supervision selection method for vision-language-action (VLA) post-training that audits predicted rewards using execution feedback. A multidimensional reward model predicts a utility score for candidate ranking and five dimensions of execution quality. An auditor uses reward predictions and uncertainty to identify overestimation patterns learned from paired imagined videos and real executions. TIF accepts the candidate with the highest predicted utility when its advantage over the existing trajectory exceeds a margin and its audit risk passes a threshold; otherwise, it retains the existing supervision. The retained real trajectories then supervise behavior cloning. Across four LIBERO suites and three training seeds, TIF achieves a mean success rate of 76.25%, improving over training on the existing trajectories by 3.00 percentage points with the same trajectory count and training budget. Adding audit gating yields 1.47 percentage points with the reward model and advantage margin held fixed. The gains persist against advantage ranking and logistic regression risk selection at matched replacement counts. On recorded videos from three real robot tasks, TIF increases clean completion among selected trajectories from 58.33% to 70.00% across 60 candidate pools.

arXiv:2611.04848v1cs.AI

Amortized Nonlinear Variable Selection with Prior-Data Fitted Networks

Variable selection is an important task for interpreting potentially complex systems in fields such as biology, business, and sociology. The importance of variable selection has attracted considerable attention in statistics and machine learning. In this paper, we are mainly interested in model-agnostic variable selection for identifying variables that have general nonlinear dependence on the response. We propose MambaPFN, short for Model-Agnostic Markov BlAnket selection with Prior-data Fitted Networks, which is built upon the Transformer, a building block of many popular large language models. Specifically, we consider a functional-mapping perspective on variable-selection learning, in which the Transformer is treated as a learnable mapping from a dataset to its relevant variables. To learn this mapping, MambaPFN introduces three new components: (i) a faithful, rich, and scalable scheme for generating pretraining data, (ii) an objective that targets each variable's posterior inclusion probability, and (iii) a modified Transformer architecture tailored to variable selection. Once pretrained, MambaPFN is applied to new datasets without parameter updates. Empirically, MambaPFN achieves desirable variable-selection performance in simulation studies and exhibits a scaling law---more pretraining datasets and larger models can continuously improve its performance. Applied to a Facebook comments dataset, MambaPFN selects the fewest variables and achieves the highest computational efficiency and predictive performance, improving upon the best baselines by about 20%.

arXiv:2611.04841v1cs.AI

DGC-Net: Cross-Modal Dynamic Routing with Global-Guided Compensation in Multimodal Sentiment Analysis

User-generated videos, including opinion vlogs and video reviews, provide a rich source of people's attitudes toward products, services, and social topics. Understanding the sentiment expressed by speakers is important for analyzing online opinions, as transcripts alone may not capture the attitude and intensity conveyed through vocal tone and facial expressions. Multimodal sentiment analysis (MSA) addresses this need by integrating textual, acoustic, and visual cues. However, these cues are often subtle, noisy, and inconsistent across modalities. Existing local affective cue modeling methods typically operate within individual modalities and overlook global contexts from other modalities. This may cause cross-modal affective cue inconsistency. Hierarchical global-local fusion methods integrate global semantics with fine-grained local affective cues. However, they commonly determine the relative contributions of the two representations implicitly, preventing the global-local contributions from adapting to the sample-dependent reliability of local cues and resulting in either insufficient use of informative local cues or excessive interference from unreliable ones. To address these issues, we propose DGC-Net, a Dynamic Gating and Global-Guided Compensation Network for MSA. DGC-Net mitigates cross-modal affective cue inconsistency through cross-modal contribution-aware dynamic expert routing. This mechanism uses auxiliary global contexts to guide local expert selection. Moreover, global-guided dynamic compensation fusion alleviates sample-dependent global-local contribution imbalance while preserving global semantic as the primary anchor and adaptively incorporating expert-refined local cues. Experiments on five multimodal benchmark datasets demonstrate the effectiveness and general applicability of DGC-Net, and ablation studies verify the contribution of each proposed component.

arXiv:2611.04838v1cs.LG

Found by Search, Broken by a Shortcut: Transfer and Tampering of Searched Intrinsic Rewards

Can a search over innate drives find intrinsic rewards that help a reinforcement-learning agent learn, transfer to a task it never saw, and resist an exploit that was absent during the search? Each candidate reward combines four generic events (a new view, a pickup, a drop and a toggle), each with a weight, a relief and a refill rate. We ran three pre-registered stages; the search protocol was locked in a public, timestamped commit before any search result existed. CMA-ES and an equal-budget random search used 1,560 PPO training runs, and the final evaluation and tamper test used 340. On the held-out task, KeyCorridorS3R2, the CMA-ES-selected reward (E) beat task-only PPO, episodic novelty and the random-search-selected reward (R) in all 3 registered comparisons, all of them meaningful by the registered size rule. Among these generic rewards E transferred best; the earlier paper's satiation reward, which on this task is given hand-wired key and locked-door events, did better (20 of 20 seeds solved against 4 of 20; descriptive). On UnlockPickup, where that reward uses generic events, E's mean area exceeded it by 0.276 (descriptive). On the training tasks there was no registered difference between E and R: both E-over-R tests missed. A free reset action, absent during the search, collapsed E's learning on DoorKey-8x8 (mean area 0.829 without it, 0.136 with it; registered) to below task-only PPO (descriptive). The searched drives helped on every task tested, but a reward given task-specific events did better on the held-out task, and selection gave E no protection against an exploit absent from the search.

arXiv:2611.04836v1cs.AI

HTCE: Homeomorphism-inspired Topological Cross-Scene Experience Canonicalization and Reuse for Object Navigation

As embodied agents perform object-goal navigation across residential environments, they accumulate successful experiences whose semantic structure recurs across scenes even when geometric trajectories do not. Reusing such experience is challenging because differences in region naming and observation granularity produce distinct graph representations of similar approaches, while meaningful differences in anchors, branches, and endpoints must stay distinguishable. We introduce HTCE, a homeomorphism-inspired topological framework that canonicalizes and reuses navigation experience. Feature and topology subdivisions characterize representational variations, and their semantically constrained reductions produce comparable semantic sequences and structural signatures while preserving semantic anchors, branching relations, and navigation endpoints. HTCE then retrieves historical successful approaches through subdivision-aware matching as guidance for a frozen vision-language policy. Canonicalization compresses the experience index losslessly, merging 17.7--23.6\% of distinct stored route descriptions while retaining full task performance. Under a single frozen vision-language backbone shared by all compared systems, HTCE attains the best SR and SPL in every setting---59.9\%/31.8\% on HM3D v0.1, 71.0\%/31.5\% on HM3D v0.2, and 46.6\%/16.6\% on MP3D---with SR gains of up to 4.9 points and structured retrieval contributing up to 2.8 SPL points. The code is available in https://github.com/startracker0/htce.git

arXiv:2611.04835v1cs.LG

Anomaly Detection on Resource-Constrained Devices with Reservoir Computing

Anomaly detection at the edge requires the deployment of models that achieve competitive performance, while minimizing computational cost and memory footprint. In this article, we investigate the suitability of the Mahalanobis Distance of Reservoir States (MDRS), an unsupervised time-series anomaly detection method, for resource-constrained devices and propose two optimizations to reduce its computational requirements. Specifically, we replace random reservoir connectivity (MDRS-Rand) with a circular shift (MDRS-Circ) and show that the anomaly score for an entire time series can be computed from the final reservoir state alone. Furthermore, we show that exploiting the reservoir's temporal memory enables more efficient input preprocessing, further reducing computational cost. We evaluate the MDRS variants on the MLPerf Tiny anomaly detection benchmark against the reference autoencoder and other anomaly detection methods, and deploy their inference kernels on a NUCLEO-F401RE microcontroller to assess latency, estimate energy consumption, and measure memory footprint. Among the evaluated methods, only MDRS and the autoencoder meet the benchmark's performance target, with MDRS requiring fewer operations and smaller models. With optimized preprocessing, MDRS achieves an ROC AUC of 0.869, a model size of 24 KB, an estimated energy consumption of 24 uJ per inference, and a latency of 0.32 ms, corresponding to a 75.9x speedup over the autoencoder with standard preprocessing.

arXiv:2611.04834v1cs.AI

Feeling of Knowing in Large Language Models

Post-training updates change what large language models (LLMs) can answer, but may leave their answerability judgments misaligned. We study this knowledge-judgment desynchronization through functional feeling of knowing (FoK), which supports selective generation, retrieval triggering, and model routing without implying subjective experience. We propose two trainable channels: a knowledge channel that changes answerability and a FoK channel that judges the current knowledge state. Three stages train them: on-policy supervised fine-tuning (OP-SFT) updates knowledge from corrected on-policy answers; knowledge-delta FoK training (K$\Delta$-FT) adapts judgments across intermediate knowledge states; and FoK-guided policy optimization (FGPO) trains the subsequent answer or explanation. Factual-QA experiments, cross-dataset evaluation, and three sequential updates show that explicit FoK alignment improves judgment quality on updated knowledge. MMLU and ARC-Easy separately assess capability preservation.

arXiv:2611.04822v1cs.CV

Radial distortion homography estimation from affine-covariant or orientation-covariant features

We present minimal solvers for joint homography and radial distortion estimation using affine- or orientation-covariant features. The core contribution is a set of novel constraints that exploit local feature geometry inherent to affine descriptors, enabling solvers that require fewer correspondences than point-based methods. Using modern robust estimation frameworks, we demonstrate that the proposed solvers are advantageous to the current state-of-the-art with a significant speed-up in the presence of outliers while maintaining comparable or superior estimation accuracy. The proposed solvers are tested on both synthetic and real images containing challenging fisheye distortions.

arXiv:2611.04819v1cs.AI

Reconstructing Fraudulent Strategies in Online Payments via Interpretable Inverse Reinforcement Learning

Payment fraud causes financial losses and undermines users' trust in web payment services. Although fraud detection is a critical defense for payment security, complex and diverse fraudulent behaviors can still evade detection. Generating fraud samples for data augmentation can broaden detectors' coverage of different fraud patterns. However, existing methods matching the distribution of transaction attributes alone do not ensure that generated sequences preserve the sequential relationships among transaction choices, payment feedback, and subsequent behavioral adjustments. To this end, we propose PaySeqGen, an interpretable inverse reinforcement learning framework that reconstructs fraud behavior policies from fraudsters' observed payment trajectories. We design a mechanism for constructing candidate actions to support plausible payment sequence generation while enabling exploration. The framework also learns interpretable behavioral rewards that capture fraudsters' preferences across different transaction contexts. By alternating between reward learning and policy optimization, the framework reconstructs policies that generate new fraud sequences for detector augmentation and learns interpretable rewards supporting sequential fraud-pattern analysis in real trajectories. Experiments in real-world payment scenarios at Ant International, a leading global digital payment provider, show that PaySeqGen outperforms state-of-the-art generative augmentation methods. Reward-based trajectory analysis also identifies fraud patterns of practical value, helping analyst locating suspicious segments and examine multistep fraud tactics. Evaluation in the deployed system of online payment further confirms its effectiveness.

arXiv:2611.04818v1cs.LG

Sleep-Disordered Breathing Detection with Quantum Recurrent Neural Networks

This work investigates the application of Quantum Recurrent Neural Networks to the binary classification of sleep-disordered breathing events in time-series data, aiming to distinguish between apneic and normal breathing episodes. A variational classical--quantum architecture is proposed and evaluated against classical Recurrent Neural Network models. Experiments were conducted on multiple real-world datasets with varying levels of complexity to assess the strengths and limitations of the quantum approach. In the comparative analysis, both classical and quantum models showed comparable performance. The quantum model achieved superior performance on the smallest dataset, while yielding slightly weaker results in the most complex setting. However, the quantum approach exhibited notable strengths, such as competitive performance despite a drastically reduced parameter count and low-qubit regime, as well as good generalization capabilities. These results suggest that, despite current hardware and algorithmic limitations, variational quantum recurrent models represent a promising alternative for time-series classification as Quantum Computing technologies continue to mature.

arXiv:2611.04817v1cs.AI

The Pen is Mightier than the Pixel: Natural Textual Unlearnable Example for Large Vision Language Models

The proliferation of Large Vision Language Models (LVLMs), trained on vast and often sensitive datasets, raised concerns regarding data privacy and unauthorized usage. Unlearnable Examples (UEs), which subtly perturb training data to mislead the models, have emerged as a promising defense method. This paper presents the first exploration of UEs specifically on LVLMs, which poses unique challenges due to their inherent multimodal nature. We propose a novel text-based perturbation method, Natural Textual Unlearnable Examples (NTUE), which can effectively mislead LVLMs by subtly manipulating the image label, establishing a shortcut between an identity of interest and a separate target, without introducing factual misinformation or affecting readability. Comprehensive experiments compared with traditional UE methods demonstrate the superior performance of NTUE. Our findings show that NTUE achieves a high success rate with a poison rate as low as 5%, underscoring its potential as a practical and robust solution for protecting multimodal data in the era of LVLMs.

arXiv:2611.04814v1cs.AI

Think Less, Locally: Local Preference Learning for Reasoning Compression

Large reasoning models achieve strong performance on challenging tasks, but their long chain-of-thought often incurs substantial inference costs by continuing to deliberate after sufficient information has been obtained. Existing reasoning compression methods primarily operate on reasoning trajectories, paying less attention to the local decision of whether further reasoning is necessary given the current state. We propose LoCo, a local preference learning method for reasoning compression. LoCo detects candidate overthinking states from the model's own traces, freezes the corresponding reasoning prefixes, regenerates shorter verified continuations from the same states, and trains on the resulting state-conditioned preference pairs with DPO. This formulation teaches the model when to continue reasoning and when to conclude, rather than simply encouraging shorter trajectories overall. We construct training data entirely from self-generated mathematical traces and evaluate LoCo on five mathematical and three code generation benchmarks. LoCo reduces completion length by 41.7\%--45.4\% on mathematics and 23.4\%--36.3\% on code, while maintaining or improving accuracy and achieving the best AUC across all code benchmarks. LoCo consistently improves the accuracy-efficiency trade-off on mathematics and transfers effectively to code generation despite math-only training, demonstrating the cross-domain generality of local preference learning for reasoning compression. The code and model will be publicly released at https://github.com/MuyaoYuan/LoCo.

arXiv:2611.04811v1cs.CV

NightIQA: A Comprehensive Dataset and Benchmark for MLLM-Based Nighttime Image Quality Assessment

Nighttime images are inherently susceptible to complex, compound degradations, making accurate quality assessment exceptionally challenging. Score-based IQA methods condense quality into a single number, lacking the multi-dimensional, language-grounded reasoning required for nighttime evaluation. Multimodal Large Language Models (MLLMs) offer a more comprehensive alternative, yet remain constrained by a critical data gap: existing real-world IQA datasets are overwhelmingly daytime-centric and often built from outdated captures, while synthetic alternatives lack nighttime-specific degradation modeling, yielding artifacts far removed from authentic nighttime conditions. To bridge this gap, we present NightIQA, a dataset and benchmark suite purpose-built for MLLM-based nighttime IQA. It comprises three core components: NightIQA-Real, a real-world dataset of 1,769 high-resolution images that we captured ourselves with modern equipment, encompassing six canonical distortion types and structured multi-dimensional annotations; NightIQA-Bench, a benchmark of 1,078 visually grounded multiple-choice questions tailored to MLLM-based nighttime IQA, organized around a four-level evaluation hierarchy spanning Perception, Technical Assessment, Localization, and Reasoning; and NightIQA-Syn, an inherently scalable, scene-aware synthesis pipeline that jointly models global-level and instance-level degradations, with an initial release of 80k high-fidelity images alongside the full generation code. Fine-tuned on NightIQA with Supervised Fine-Tuning (SFT) and Group Relative Policy Optimization (GRPO), Qwen3-VL-8B outperforms open-source IQA methods and achieves higher overall accuracy than leading closed-source models, highlighting the value of domain-specific data and structured supervision.

arXiv:2611.04809v1cs.AI

RAHC: Rule-Augmented Hyper-Relational Knowledge Graph Completion for Materials Synthesis

Scientific discovery depends on reusing experimental knowledge, yet records scattered across papers and repositories are often incomplete. In materials synthesis, recovering missing conditions requires reasoning about the target material and the observed context of a specific experiment. Hyper-relational knowledge graphs (HKGs) represent this context through qualifiers, but existing completion methods leave domain knowledge and experimental regularities implicit. We propose RAHC, a rule-augmented HKG framework for single-slot and joint multi-slot synthesis-condition completion. RAHC builds a traceable rule catalog from training records and retrieved literature, retrieves rules for each query-candidate pair, and separately encodes graph structure and rule evidence. Hard rules constrain candidate validity, while soft rules guide ranking. RAHC-G aggregates rules using global reliability; RAHC-Q additionally learns query-specific relevance. Gated fusion combines structural and rule representations, while tuple-conclusion rules provide cross-condition evidence for joint completion. Experiments on five materials-synthesis datasets show improvements in single-slot and joint completion under Semi-inductive and Open-Answer Inductive settings, with Semi-inductive MRR gains of up to 15.6% over MAYPL. RAHC also improves performance on the entity-disjoint ILPC'22 Small benchmark, supporting transfer to fully inductive KG completion. These results support experimental condition selection, synthesis planning, and the reuse of scientific knowledge.

arXiv:2611.04807v1math.OC

Asymptotic Optimality of $c\mu$-Rule for LLM Inference Scheduling under Memory Constraint

Large language model (LLM) inference, the process by which a pre-trained model receives an input prompt and generates output tokens sequentially, has become increasingly important with the widespread deployment of LLM-based applications. As inference workloads continue to scale, GPU memory has emerged as a critical bottleneck: the long input prompts and generation sequences common in agentic and reasoning workloads lead to substantial key-value (KV) cache usage. In this paper, we study the resulting online scheduling problem from a queueing perspective. Requests arrive stochastically with heterogeneous prompt and output lengths, undergo prefill and autoregressive decode, and compete for a finite KV-cache capacity. Unlike classical queueing models, the duration of a processing step depends on both prefill computation and KV-cache I/O, so service capacity is jointly determined by computation, memory access, and memory occupancy. We develop a fluid model that reduces the system to a multiclass single-server queue with effective service rates capturing computation, memory access, and memory capacity. This fluid analysis yields a $c\mu$ priority rule, where the effective service rate $\mu$ is jointly determined by computational speed, memory bandwidth, and memory capacity. We prove that this rule minimizes cumulative holding cost in the fluid model and, for a finite workload, also minimizes completion time and maximizes full-run output-token throughput. We then construct a feasible scheduling policy for the original finite system and prove that its scaled dynamics converge to the optimal fluid policy. Finally, experiments on a real LLM serving engine using an NVIDIA H800 show that practical implementations of the proposed priority rule substantially reduce average latency while maintaining high throughput relative to existing scheduling policies.

arXiv:2611.04806v1cs.LG

On the Complexity of Finite-Sum Nonconvex-Strongly-Concave Minimax Optimization with Exact Proximal Oracles

Finite-sum nonconvex-strongly-concave minimax problems arise in adversarial training and distributionally robust learning. Their known lower bounds cover only linear-span algorithms, zero-respecting algorithms, or a restricted proximal protocol. They leave open the power of exact component proximal steps, whose responses are global saddle points of regularized component problems and can expose directions that gradients do not reveal. We show that neither unrestricted adaptivity nor exact proximal access improves the polynomial complexity. We consider $n$ components with joint mean-squared smoothness $L$ and condition number $\kappa$, started from a point with primal gap at most $\Delta$ and joint gradient norm at most $\sqrt{2L\Delta}$. For $\varepsilon^2/(L\Delta)$ below a universal constant, every randomized adaptive algorithm with exact component values, gradients, and proximal responses needs $\Omega(n+\min\{\sqrt n\,\kappa,n^{3/4}\sqrt\kappa\}L\Delta/\varepsilon^2)$ component queries in the worst case to find an $\varepsilon$-stationary point of the value function. A method using only component gradients attains this rate up to a factor $\log^2(2+\kappa)$, including initialization. Zhou (arXiv:2610.04944) concurrently proves the same rate for zero-respecting first-order methods. In our hard instances, every component admits proximal steps with the same step sizes, and a response for a component reveals hidden directions only up to the first one assigned to another component.

arXiv:2611.04805v1cs.AI

Where Is the Truth When AI Agents Disagree? Dependence-Aware Belief Propagation for Unsupervised Truth Discovery (Extended Version)

LLM agents often give conflicting answers to the same question, and many deployed agents are variants of one base model, so their errors are correlated. We study truth discovery among such agents: inferring the true answer and each agent's reliability without labels. Classical truth-discovery and crowdsourcing estimators reward an agreeing bloc and can fall below majority vote. We propose BP-TD, which adds per-agent decorrelation weights to mean-field variational inference on a one-coin Dawid-Skene model, so that agents sharing a base model count together as about one vote. Its first iteration is exactly a weighted vote; later iterations estimate agent reliability. With 13 local agents on MMLU, BP-TD beats majority vote without any labels (0.716 vs. 0.624 on 160 questions; 0.707 vs. 0.674 on 10,000) and ranks agents by their true accuracy, which classical estimators invert. The gain depends on the pool: over 2,852 sub-pools of four benchmarks, decorrelation helps when the largest correlated bloc is weaker than the other agents and hurts when it is the strongest, and we found no label-free statistic that tells the two cases apart. Labels change the picture quickly: with 20 labelled questions, selecting the single best agent is more accurate than every aggregator we tested on MMLU, ARC and TruthfulQA.