arXiv:2605.02277v2 Announce Type: replace
Abstract: Factual Error Correction (FEC) aims to revise inaccurate text into statements that are factually consistent with external evidence. Although recent methods perform well on single-hop correction, they often treat claims as atomic units and struggle with multi-hop cases that require compositional reasoning across multiple evidence sources. This challenge is further amplified by limited paired data and difficulties in locating semantic errors within complex reasoning chains. We present CECoR (Compositional Error Correction via Reasoning-aware Synthesis), a reasoning-aware framework that introduces a Decomposition and Injection paradigm for compositional error correction. CECoR decomposes multi-hop claims into interpretable reasoning steps and injects controlled perturbations to synthesize high-quality training pairs. A two-stage learning strategy combining supervised fine-tuning and reinforcement learning improves factual accuracy and robustness. Comprehensive evaluations show that CECoR achieves strong performance on multi-hop benchmarks, outperforming both distantly supervised methods and few-shot LLM baselines. It also generalizes effectively to single-hop correction and remains stable under noisy evidence, demonstrating its versatility for real-world factual correction.
Science Journals
arXiv:2605.03052v2 Announce Type: replace
Abstract: We study how Large Language Models (LLMs) process negation mechanistically. First, we establish that even though open-weight models often provide wrong answers to questions involving negation, they do possess internal components that process negation correctly. Their poor accuracy is due to late-layer attention behavior that promotes simple shortcuts; ablating those attention modules greatly improves accuracy on negation-related questions. Second, we uncover how models process negation. We consider two hypotheses: models could use attention heads that attend to the phrase being negated and suppress related concepts, or they could directly construct a representation of the entire negative phrase (e.g., representing "not gas" as a vector that promotes liquids and solids). We apply a range of observational and causal interpretability techniques on Mistral-7B and Llama-3.1-8B to show that models implement both mechanisms, with the "constructive" mechanism being more prominent. Combined, our work deepens the understanding of LLMs' internals, highlighting construction-dominant computations and the coexistence of competing mechanisms within LLMs.
arXiv:2605.03384v2 Announce Type: replace
Abstract: Acoustic side-channel attacks (ASCA) on keyboards pose a significant security risk, as keystrokes can be inferred from typing acoustics, revealing sensitive information. Prior ASCA studies are limited by small-scale datasets with restricted diversity in users, keyboards, and environments, constraining analysis across devices, microphones, and noise conditions. We introduce HEAR, a dataset designed to study ASCA along three axes: keyboard generalization, noise adaptation, and user bias. HEAR contains recordings from 53 participants using 37 laptop keyboards, collected in three realistic settings: (1) external microphone capture, (2) device microphone capture without network noise, and (3) VoIP-based streaming capture. This enables controlled evaluation across users, keyboards, and environments. On HEAR, we establish an ASCA benchmark spanning conventional features and pre-trained representations from raw audio and spectrograms in unimodal and multimodal settings. We propose DECKER, a domain-invariant keystroke inference framework with four stages: (1) Keyboard Signature Normalization to reduce device coloration, (2) domain-adversarial disentanglement to suppress keyboard identity, (3) supervised cross-keyboard contrastive alignment to enforce key consistency, and (4) Acoustic Style Randomization to synthesize unseen keyboard responses. We further explore sentence-level inference using an LLM-based post-processing layer to refine keystroke sequences via linguistic context. Results on HEAR show DECKER improves keystroke identification over strong baselines, particularly in cross-keyboard and cross-user settings, with further gains from language-model rectification. These findings highlight that ASCA remains effective across diverse users, devices, and noisy environments, underscoring its practical security risk.
arXiv:2605.03403v2 Announce Type: replace
Abstract: Group Relative Policy Optimization (GRPO) has recently shown strong performance in post-training large language models and vision-language models. It raises a question of whether the GRPO also significantly promotes the test-time adaptation (TTA) of vision language models. In this paper, we propose Group Relative Policy Optimization for Test-Time Adaptation (GRPO-TTA), which adapts GRPO to the TTA setting by reformulating class-specific prompt prediction as a group-wise policy optimization problem. Specifically, we construct output groups by sampling top-K class candidates from CLIP similarity distributions, enabling probability-driven optimization without access to ground-truth labels. Moreover, we design reward functions tailored to test-time adaptation, including alignment rewards and dispersion rewards, to guide effective visual encoder tuning. Extensive experiments across diverse benchmarks demonstrate that GRPO-TTA consistently outperforms existing test-time adaptation methods, with notably larger performance gains under natural distribution shifts.
arXiv:2605.04127v2 Announce Type: replace
Abstract: Model collapse, the degradation in performance that arises when generative models are trained on the outputs of prior models, is an increasing concern as artificially generated content proliferates. Related critiques of large language models have highlighted their tendency to reproduce frequent patterns in training data, their reliance on vast datasets, and their substantial environmental cost. Together, these factors contribute to data degradation, the reinforcement of cultural biases, and inefficient resource use. In this position paper we aim to combine these views and argue that model collapse threatens current efforts to democratize AI. By reducing training efficiency and skewing data distributions away from the tails of their support, model collapse disproportionately impacts low-resource and marginalized communities. We examine both the environmental and cultural implications of this phenomenon, situate our position within recent position papers on model collapse, and conclude with a call to action. Finally, we outline initial directions for mitigating these effects.
arXiv:2605.04193v2 Announce Type: replace
Abstract: Inductive Logic Programming (ILP) aims to learn interpretable first-order rules from data, but existing symbolic and neuro-symbolic approaches struggle to scale to noisy and probabilistic settings. Classical ILP relies on discrete combinatorial rule search and is brittle under uncertainty, while differentiable ILP methods typically depend on predefined rule templates or inaccurate fuzzy operators that suffer from vanishing gradients or poor approximation of logical structure when reasoning over probabilistic predicate valuations. This paper proposes an Attention-based Neuro-symbolic Differentiable Rule Extractor (ANDRE), a novel ILP framework that learns first-order logic programs by optimizing over a continuous rule space with attention-based logical operators. ANDRE replaces both rule templates and logical operators with fully differentiable, attention-driven conjunction and disjunction operators that approximate logical min-max semantics, enabling accurate, stable, and interpretable reasoning over probabilistic data. By softly selecting, negating, or excluding predicates within each rule, ANDRE supports flexible rule induction while preserving symbolic structure. Extensive experiments on classical ILP benchmarks, large-scale knowledge bases, and synthetic datasets with probabilistic predicates and noisy supervision demonstrate that ANDRE achieves competitive or superior predictive performance while reliably recovering correct symbolic rules under uncertainty. In particular, ANDRE remains robust to moderate label noise, substantially outperforming existing differentiable ILP methods in both rule extraction quality and stability.
arXiv:2605.05427v2 Announce Type: replace
Abstract: Refusal rates are a poor proxy for LLM safety, i.e., a model may over-refuse benign prompts while still complying with harmful ones. We audit both failure modes across 21 open-weight LLMs on four safety benchmarks (OR-Bench, XSTest, ToxiGen, BOLD), using a composition adjustment to isolate model sensitivity from dataset toxicity confounds. We report three findings. First, models adopt fundamentally different calibration strategies: conservative ecosystems such as Llama suppress unsafe outputs at the cost of elevated over-refusals, while permissive ecosystems such as DeepSeek and Qwen preserve helpfulness but tolerate higher harmful compliance. Second, demographic protection is unequal: models over-protect prominent racial and religious groups, frequently refusing even benign prompts about them, while providing substantially weaker protection against disability-targeted attacks. Third, refusal and compliance tendencies are stable within model families across generations and scales, suggesting that post-training objectives shape safety behavior more than architecture. Our results call for joint, demographically-aware, and multi-judge safety evaluation.
arXiv:2605.05713v2 Announce Type: replace
Abstract: Infrared breast thermography provides a noninvasive measurement of skin-surface temperature, but the relation between surface thermal patterns and intratumoral physiology is limited by heat diffusion and thermal screening. Here we study a steady-state modified Pennes bioheat model in a two-dimensional multilayer breast-tissue cross-section containing a finite-sized tumor with spatially heterogeneous perfusion. We compare four idealized perfusion patterns: uniform, rim-enhanced, necrotic-core, and anisotropic perfusion. To assess how well these internal differences are preserved at the surface, we compare the full temperature-rise profiles using an $L^2$ distance and define thermal-signature equivalence through an observational tolerance. The results show that distinct perfusion patterns can generate clearly different internal temperature fields, while their surface signatures may become much more similar after propagation through the surrounding tissue. Tests with noisy surface profiles indicate that this equivalence classification is sensitive to the assumed form of profile-level uncertainty. After matching the tumor-averaged perfusion, the radially heterogeneous cases become much closer to the uniform case, whereas the anisotropic case remains more distinguishable because of its directional structure. Increasing tumor depth promotes thermal-signature equivalence, whereas increasing tumor diameter enhances surface distinguishability;a depth--diameter map shows the competition between these two effects. Fat-layer thickness and mild outer-surface deformation modify the surface profiles, but their influence is secondary over the parameter ranges considered here. These results highlight a limitation of static breast thermography: a surface thermal anomaly can be detected without uniquely identifying the underlying intratumoral perfusion structure.
arXiv:2605.07527v2 Announce Type: replace
Abstract: Recent work has observed that explanations produced by Self-Interpretable Graph Neural Networks (SI-GNNs) can be self-inconsistent: when the model is reapplied to its own explanatory graph subset, it may produce a different explanation. However, why self-inconsistency arises remains poorly understood. In this work, we first identify re-explanation-induced context perturbation as the direct cause of score variation. We then introduce a latent signal assignment hypothesis to explain why only some edges are sensitive to this perturbation, and analyze how conciseness regularization affects latent signal assignment. Given that self-inconsistent edges do not provide stable evidence for the model's prediction, we propose Self-Denoising (SD), a model-agnostic and training-free post-processing strategy that calibrates explanations with only one additional forward pass. Experiments across representative SI-GNN frameworks, backbone architectures, and benchmark datasets support our hypothesis and show that SD consistently improves explanation quality while adding only about 4--6\% computational overhead in practice.
arXiv:2606.01372v1 Announce Type: new
Abstract: Can neural networks learn abstract algebraic rules, or do they merely memorize training patterns? We investigate this using MNIST digits as states and modular arithmetic operations as actions in a JEPA-style latent world model. Standard supervised baselines and JEPA models with additive operation embeddings fit seen operations but fail to extrapolate reliably to unseen ones. To bridge this gap, we introduce a block-rotation predictor that imposes the circular structure of modulo-10 arithmetic in latent space. This enables strong zero-shot generalization, with the best ResNet-based JEPA block-rotation model achieving 99.46\% zero-shot and 99.46\% rollout accuracy. Our results suggest that latent world models can learn symbolic transformation rules when architecture matches the structure of the problem. Our code can be \href{https://github.com/DL-World-Models/mnist-math}{accessed here}.
arXiv:2602.14878v3 Announce Type: replace
Abstract: The Model Context Protocol (MCP) introduces a standard specification that defines how Foundation Model (FM)-based agents should interact with external systems by invoking tools. However, to understand a tool's purpose and features, FMs rely on natural-language tool descriptions, making these descriptions a critical component in guiding FMs to select the optimal tool for a given (sub)task and to pass the right arguments to the tool. While defects or smells in these descriptions can misguide FM-based agents, their prevalence and consequences in the MCP ecosystem remain unclear.
Hence, we examine 856 tools spread across 103 MCP servers empirically, assess their description quality, and their impact on agent performance. We identify six components of tool descriptions from the literature, develop a scoring rubric utilizing these components, and then formalize tool description smells based on this rubric. By operationalizing this rubric through an FM-based scanner, we find that 97.1% of the analyzed tool descriptions contain at least one smell, with 56% failing to state their purpose clearly. While augmenting these descriptions for all components improves task success rates by a median of 5.85 percentage points and improves partial goal completion by 15.12%, it also increases the number of execution steps by 67.46% and regresses performance in 16.67% of cases. These results indicate that achieving performance gains is not straightforward; while execution cost can act as a trade-off, execution context can also impact. Furthermore, component ablations show that compact variants of different component combinations often preserve behavioral reliability while reducing unnecessary token overhead, enabling more efficient use of the FM context window and lower execution costs.
arXiv:2605.15746v5 Announce Type: replace
Abstract: Privacy-preserving cryptocurrency exchanges alter what the pricing mechanism observes about order flow. We derive the unique linear Kyle equilibrium when a committed Bayesian market maker observes order flow perturbed by independent Gaussian privacy noise. The price-impact coefficient and informed-trader strategy rescale by reciprocal factors of the privacy parameter (one down, one up), so their product is invariant. A welfare decomposition then identifies a closed-form per-period transfer from the protocol's LP pool to traders -- the "privacy subsidy", the break-even fee any privacy-aggregated exchange must charge. The result is the single-period closed-form privacy-noise analog of Loss-Versus-Rebalancing (Milionis et al. 2022). The primary application is shielded AMMs with explicit additive-noise injection (e.g., differential privacy); related designs (batched swaps, sealed-bid auctions, oracle-pegged crossings) require separate frameworks that we leave to future work.
arXiv:2605.16415v3 Announce Type: replace
Abstract: The creativity of diffusion models refers to their ability to generate highly realistic images that are different from their training data. Creativity is somewhat surprising since it is known that if the denoiser used in the diffusion model is the Bayes optimal denoiser for a given training set, then the model will simply copy the training samples. In this paper we present empirical and theoretical results that suggest that creativity in diffusion models is due to an interaction between the denoiser architecture and the target distribution. Theoretically, we give explicit forms for the distribution of generated samples as a function of the target distribution and the denoiser architecture for three different denoiser architectures (linear, polynomial, bottleneck). Empirically, we show that small changes in the popular UNET denoiser architecture leads to very different forms of creativity, and these small changes often yield samples that are highly nonrealistic. Taken together, our results show that diffusion models will only be successful if the inductive bias of the denoiser architecture is in strong alignment with the true target distribution.
arXiv:2605.16446v2 Announce Type: replace
Abstract: Semi-supervised learning (SSL) enables prediction with limited labels, but high-stakes tabular applications (medical, credit, recidivism) require statistical fairness guarantees. We identify a structural conflict in tabular fair SSL through a diagnostic stress test: under confidence-gated pseudo-labeling, moment-matching fairness regularizers can trigger two failure modes -- Masking Collapse (fairness erodes confidence, starving pseudo-labels) and Trivial Saturation (drift to constant predictors). We propose Online Primal-Dual Allocation (OPDA), an online controller that schedules fairness and entropy-based stability penalties using violation, risk, and pseudo-label health signals, avoiding per-dataset selection of a fixed fairness weight within this diagnostic regime. On the evaluated tabular benchmarks (Adult, ACSIncome, COMPAS), OPDA mitigates the degenerate regimes observed under static weighting and simple single-signal adaptive baselines. On Adult and COMPAS, it yields non-degenerate operating points competitive with the empirical static-$\lambda$ frontier; on ACSIncome, it preserves utility with a wider fairness-utility spread. Relative to OPDA-lite, the full controller mainly shifts the operating point toward higher utility on ACSIncome, while Adult highlights the fairness-utility trade-off between the two variants. These results position OPDA as a calibration-free controller for non-degenerate operating points in tabular fair SSL without per-dataset tuning.
arXiv:2605.16451v2 Announce Type: replace
Abstract: Macro placement is a pivotal stage in VLSI physical design, fundamentally determining the overall chip performance. Recent data-driven placement methods have demonstrated significant potential, yet they often struggle to handle sequential dependencies and to balance topological connectivity with physical constraints. To bridge this gap, we propose MacroDiff+, a physics-guided geometric diffusion framework. Specifically, we design a dual-domain denoising architecture that couples topological connectivity encoded by heterogeneous GNNs with global geometric context modeled by a Transformer. Furthermore, we introduce Physics-Guided Sampling, an inference strategy that actively steers the generation using explicit gradients to ensure both statistical plausibility and physical validity. On the ISPD2005 MMS benchmarks, MacroDiff+ outperforms state-of-the-art baselines with a 6.1-6.2% reduction in wirelength. Notably, it exhibits superior stability and scalability on large-scale designs where prior methods fail to converge. The source code is available at https://github.com/jhy00n/MacroDiff-plus.
arXiv:2605.16564v2 Announce Type: replace
Abstract: Discontinuous coefficient fields arise in many computational physics problems and are often represented as cellwise constant data tied to a given spatial discretization. Such representations are inherently mesh-dependent, requiring interpolation or projection whenever they are transferred to a different discretization. In this work, we develop \emph{Parallel and Adaptive Mesh-Free Approximation (PAM)}, a mesh-independent framework that approximates discontinuous data by a continuous, closed-form function. The resulting approximation can be evaluated consistently across different geometries and numerical discretizations, while preserving sharp interface features.
The proposed PAM framework employs radial basis functions (RBFs) to construct continuous approximations of discontinuous data. To accurately capture discontinuities, we incorporate Shepard-normalization, which stabilizes the approximation near sharp interfaces. The coefficients of the RBF expansion are determined via sparse regression, enabling automatic selection of the most relevant basis functions and promoting robust representations. In addition, we develop a novel adaptive refinement approach which further enriches the approximation in regions of rapid spatial variation. We provide a theoretical analysis showing that the proposed normalized RBF framework achieves arbitrarily small $L^1$ error in approximating discontinuous step functions.
To enhance computational efficiency, the domain is partitioned into subdomains, and the reconstruction problem is solved independently on each subdomain in parallel. Numerical experiments demonstrate the accuracy, adaptivity, scalability, and downstream impact of the proposed method on Darcy flow simulations, including tests on heterogeneous permeability fields, mesh-transfer settings, and the SPE10 benchmark.
arXiv:2605.16740v2 Announce Type: replace
Abstract: Multi-video event understanding demands models that can locate and attribute query-relevant evidence scattered across long, heterogeneous video corpora. Existing large vision-language models (LVLMs) often underperform in this regime because they quickly exhaust their context budget and struggle to precisely localize evidentially important segments, frequently missing dense informational cues such as broadcast graphics, subtitles, and scoreboards. We introduce TRACE, an evidence grounding-guided framework that follows a ground-before-reasoning strategy for multi-video event reasoning. Our approach first builds a structured, text-searchable timeline for each video using OCR and object detection. A text-only LLM then conducts query-aware evidence localization, selecting relevant moments prior to any downstream visual reasoning. The retrieved frames and their grounding summaries are subsequently used to steer LVLM-based claim generation and cross-video citation consolidation. Experiments on MAGMaR 2026 and WikiVideo demonstrate that structured grounding markedly boosts factual completeness and attribution fidelity. On the MAGMaR validation split, TRACE raises macro-average MiRAGE F1 from 0.705 to 0.811 compared to an unguided Qwen3-VL-30B baseline, with especially strong improvements in citation recall from 0.440 to 0.628. The method also attains state-of-the-art results on the official MAGMaR 2026 leaderboard. Code is released at https://github.com/pengyu965/TRACE.
arXiv:2605.17109v3 Announce Type: replace
Abstract: In recent years, Muon has emerged as the dominant method for training large language models, and transformers more broadly. The essential difference, when compared to standard gradient descent methods, is to replace the usual update matrix $M=U\Sigma V^\top$ with its polar factor $UV^\top$. In this work, we consider a class of Muon-like updates, where we replace the update $M$ with $U\Sigma^p V^\top$ for some parameter $p$. We call this a "spectral-shaping" operation, and develop a theory of how to pick $p$ which depends on (a) local curvature of the loss function, (b) noise stemming from stochastic gradients and label noise, and (c) training stage. Our theory and experimentation reveal a previously overlooked behavior: positive $p$ helps early by emphasizing high-curvature directions and accelerating signal contraction, while mildly negative $p$ helps later by reallocating update strength toward low-curvature directions that still contain useful training signals. Building on the insight, we propose DynMuon, an efficient dynamic spectral shaping method that schedules $p$ from positive to mildly negative over training. Extensive experiments across model sizes, architectures, and training settings show that DynMuon consistently achieves lower validation loss than Muon, while requiring 10.6-26.5% fewer steps to reach the same target loss. Our code is available at https://github.com/fzwark/DynMuon.
arXiv:2605.17110v2 Announce Type: replace
Abstract: Query clustering organizes queries into groups that reflect shared latent capability demands, enabling capability-aware LLM evaluation. Existing clustering methods, which primarily rely on semantic taxonomies or embeddings, often fail to capture such latent capability requirements due to a misalignment between surface-level semantics and actual model performance. We propose ECC, an algorithm that calibrates prior semantic embeddings using limited posterior model comparisons to bridge the gap between surface-level semantics and latent capability requirements. ECC characterizes each cluster through a capability profile parameterized by a Bradley-Terry model and uses trainable mixture weights to accommodate queries with mixed capability demands, jointly learning a flexible, capability-aware clustering structure that supports query-specific inference of LLM capabilities. Extensive quantitative and qualitative evaluations demonstrate that ECC significantly improves LLM capability ranking quality, outperforming human-labeled and embedding-based baselines by an average of 17.64 and 18.02 percentage points, respectively, and proves effective in downstream tasks such as query routing.
arXiv:2605.17921v2 Announce Type: replace
Abstract: Streaming video requires handling dynamic information density under strict latency budgets. Yet, existing methods typically employ static strategies, such as fixed memory compression or reliance on a single model, forcing a trade-off: fast models fail on complex queries, while always-on heavy models violate real-time constraints and overcomplicate simple queries. Rather than fixing these decisions upfront, we propose R3-Streaming (Remember, Respond, Reason), which formulates streaming video understanding as a cascaded control problem: for each query, the system compresses memory, judges response readiness, and routes computation sequentially, so that each downstream decision builds on progressively refined information states. To optimize this pipeline, we introduce an age-aware forgetting policy for memory compression, as aggressively compressing historical frames can yield substantial performance gains. For compute routing, we propose TB-GRPO, a target-balanced reinforcement learning objective that routes hard queries to a stronger model while preventing mode collapse. Extensive evaluations demonstrate that R3-Streaming achieves state-of-the-art results among streaming MLLMs, reaching 57.92 on OVO-Bench and 76.36 on StreamingBench, while reducing visual token usage by 95 to 96 percent.
arXiv:2605.18077v2 Announce Type: replace
Abstract: Communication is a key component in multi-agent reinforcement learning (MARL) for mitigating partial observability, yet prior approaches often rely on inefficient information exchange or fail to transmit sufficient state information. To address this, we propose LLM-driven Multi-Agent Communication (LMAC), which leverages an LLM's reasoning capability to design a communication protocol that enables all agents to reconstruct the underlying state as accurately and uniformly as possible. LMAC iteratively refines the protocol using an explicit state-awareness criterion, improving state recovery while narrowing differences in agents' knowledge. Experiments on diverse MARL benchmarks show that LMAC improves state reconstruction across agents and yields substantial performance gains over prior communication baselines.
arXiv:2605.13527v3 Announce Type: replace
Abstract: Reusable skills have become a core substrate for improving agent capabilities, yet most existing skill packages encode reusable behavior primarily as textual prompts, executable code, or learned routines. For visual agents, however, procedural knowledge is inherently multimodal: reuse depends not only on what operation to perform, but also on recognizing the relevant state, interpreting visual evidence of progress or failure, and deciding what to do next. We formalize this requirement as multimodal procedural knowledge and address three practical challenges: (I) what a multimodal skill package should contain; (II) where such packages can be derived from public interaction experience; and (III) how agents can consult multimodal evidence at inference time without excessive image context or over-anchoring to reference screenshots. We introduce MMSkills, a framework for representing, generating, and using reusable multimodal procedures for runtime visual decision making. Each MMSkill is a compact, state-conditioned package that couples a textual procedure with runtime state cards and multi-view keyframes. To construct these packages, we develop an agentic trajectory-to-skill Generator that transforms public non-evaluation trajectories into reusable multimodal skills through workflow grouping, procedure induction, visual grounding, and meta-skill-guided auditing. To use them, we introduce a branch-loaded multimodal skill agent: selected state cards and keyframes are inspected in a temporary branch, aligned with the live environment, and distilled into structured guidance for the main agent. Experiments across GUI and game-based visual-agent benchmarks show that MMSkills consistently improve both frontier and smaller multimodal agents, suggesting that external multimodal procedural knowledge complements model-internal priors.
arXiv:2605.18838v3 Announce Type: replace
Abstract: Scaling laws predict loss from compute but not how capabilities interact. We measure the coupling between reasoning and truthfulness across 63 base models from 16 families and find a regime change invisible to loss curves: below a family-dependent critical scale N_c, capabilities anticorrelate (r = -0.989, p = 4 x 10^{-5} nonparametric permutation test); above it, they cooperate. N_c ~ 3.5B parameters [2.9B, 13.4B] (bootstrap 95% CI), but model size is not the only variable that determines phase. Architecture, data curation, and training recipe each shift N_c independently: curated training eliminated the coupling dip between Qwen generations (0.025 to 0.830 at matched scale), Gemma-4 at 4B achieves coupling 0.871, characteristic of 13B+ standard-trained models, through distillation and architectural innovation, and Phi at 1B matches web-trained coupling at 10B through data curation alone. Width normalization eliminates the anticorrelation across all tested families, supporting an output-projection bottleneck. Internally, 38 of 40 models show zero competing attention heads. A sparse-regression ODE cross-predicts held-out Llama-2 at 5.6% error. The diagnostic requires no model internals -- only public benchmark scores across a model family. The cooperative regime extends to the frontier (r = +0.72, 34 models, 10 labs). A proof-of-concept intervention confirms the bottleneck is exploitable: adding a single truth-direction vector at the identified layer corrects 60% of misaligned outputs in the tax phase with zero retraining -- a surgical, per-inference correction that requires no weight modification. Code, data, an open-source steering CLI for any open-weight model, and an interactive dashboard for phase diagnosis are released: https://zehenlabs.com/cape/.
arXiv:2605.19065v3 Announce Type: replace
Abstract: In this paper, we introduce Geometric Algebra-Informed 3D Gaussian Splatting (GAI-GS), a framework for wireless modeling that couples 3D Gaussian splatting with a geometric algebra-based attention mechanism to explicitly model ray-object interactions in complex propagation environments. GAI-GS encodes joint spatial-electromagnetic (EM) relations into token representations, enabling scene-level aggregation within a unified, end-to-end neural architecture. This design grounds wireless ray propagation in electromagnetic principles, allowing token interactions to model key effects such as multipath, attenuation, and reflection/diffraction. Through extensive evaluations on multiple real-world indoor datasets, GAI-GS consistently surpasses current baselines across various wireless tasks.
arXiv:2605.19344v2 Announce Type: replace
Abstract: Linguistic cues such as "I believe" and "probably" offer an intuitive interface for communicating confidence, yet a generalisable, principled calibration framework for linguistic confidence expressions remains underexplored. In particular, co-occurring linguistic cues, contextual variation, and subjective audience interpretation pose unique challenges. We therefore model linguistic confidence as a distribution over plausible perceived probability values that a statement is correct, capturing interpretation variability that scalar representations discard. Within this distributional framework, we introduce faithfulness as a complementary evaluation dimension and present Faithfulness Divergence (FD), an information-theoretic metric quantifying the surprise induced in audience beliefs upon truth revelation. Building on these foundations, we present Retrieval-Augmented Linguistic Calibration (RALC), a lightweight post-hoc pipeline that propagates calibrated confidence signals back into natural language via retrieval-augmented rewriting. Across three QA benchmarks and five LLM families, RALC improves in-domain faithfulness and calibration up to 66% and 58%, respectively, outperforming black-box and grey-box calibration baselines.