Evaluating LLM forecasting capabilities is constrained by a fundamental tension: prospective evaluation offers methodological rigor but prohibitive latency, while retrospective forecasting (RF)—evaluating on already-resolved events—faces rapidly shrinking clean evaluation data as SOTA models possess increasingly recent knowledge cutoffs. Simulated Ignorance (SI), prompting models to suppress pre-cutoff knowledge, has emerged as a potential solution. We provide the first systematic test of whether SI can approximate True Ignorance (TI). Across 470 competition-level questions and 9 models, we find that SI fails systematically: (1) cutoff instructions leave a 52% performance gap between SI and TI; (2) chain-of-thought reasoning fails to suppress prior knowledge, even when reasoning traces contain no explicit post-cutoff references; (3) reasoning-optimized models exhibit worse SI fidelity despite superior reasoning trace quality. These findings demonstrate that prompts cannot reliably "rewind" model knowledge. We conclude that RF on pre-cutoff events is methodologically flawed; we recommend against using SI-based retrospective setups to benchmark forecasting capabilities.
论文检索
输入标题、作者或关键词，从 7,537 篇学术成果中精准定位
Unlocking the spatial intelligence of multimodal large language models (MLLMs) is crucial for understanding and interacting with the 3D world. Prevailing approaches typically inject spatial priors via external tools, which impose significant inference overhead, or rely on latent feature distillation, which remains uninterpretable and lacks fine-grained geometric constraints. To address these issues, we propose SpatialSV, a framework designed to internalize robust 3D spatial awareness within MLLMs while simultaneously offering inherent interpretability. Deviating from passive feature imitation, SpatialSV employs task-oriented visual supervision, compelling the model to actively lift its 2D visual features into explicit 3D representations, including depth maps, camera poses, and point clouds. Crucially, this 2D-to-3D lifting process provides a transparent window into the model’s representations: the resulting 3D reconstructions serve as an intuitive proxy for visualizing and diagnosing the quality of the model’s intrinsic spatial knowledge. Extensive experiments across multiple benchmarks and models demonstrate the effectiveness of SpatialSV in enhancing and interpreting MLLMs’ spatial intelligence. Furthermore, the framework exhibits strong generalization in semi-supervised settings, validating its potential to leverage unlabeled visual data for scalable, interpretable spatial representation learning.
Vision-language model (VLM) encoders such as CLIP enable strong retrieval and zero-shot classification in a shared image–text embedding space, yet the semantic organization of this space is rarely inspected. We present a post-hoc framework to explain, verify, and align the semantic hierarchies induced by a VLM over a given set of child classes. First, we extract a binary hierarchy by agglomerative clustering of class centroids and name internal nodes by dictionary-based matching to a concept bank. Second, we quantify plausibility by comparing the extracted tree against human ontologies using efficient tree- and edge-level consistency measures, and we evaluate utility via explainable hierarchical tree-traversal inference with uncertainty-aware early stopping (UAES). Third, we propose an ontology-guided post-hoc alignment method that learns a lightweight embedding-space transformation, using UMAP to generate target neighborhoods from a desired hierarchy. Across 13 pretrained VLMs and 4 image datasets, our method finds systematic modality differences: image encoders are more discriminative, while text encoders induce hierarchies that better match human taxonomies. Overall, the results reveal a persistent trade-off between zero-shot accuracy and ontological plausibility and suggest practical routes to improve semantic alignment in shared embedding spaces.
Knowledge editing is pivotal for efficiently updating the parametric memory of Large Language Models (LLMs), enabling them to function as evolving agents in dynamic environments. However, mainstream in-parameter knowledge editing approaches suffer from Subject-Dominant Memory Interference: modifying a specific fact inadvertently corrupts the broader structural knowledge associated with the same subject within LLMs. We diagnose the root cause as a shortcut learning pathology, where the optimization objective overfits subject representations while bypassing the essential relational context. To rectify this, we propose Causal Path Alignment (CPA), a principled framework designed to anchor the optimization trajectory to valid causal pathways. CPA enforces parameter updates to route through relation-aware intermediate states, thereby preventing the erasure of contextual dependencies. Experimental results across diverse LLM backbones demonstrate that CPA consistently eliminates the shortcut, significantly improving relation specificity while exhibiting minimal side-effects. Moreover, CPA serves as a model-agnostic plug-in for existing editors, paving the way for reliable and trustworthy in-parameter knowledge editing.
When vision contradicts text, multimodal large language models (MLLMs) consistently favor text—even when images provide clear evidence otherwise. This bias poses risks for applications requiring visual grounding, yet its cause remains unclear. In this paper, we uncover a surprising finding: models often get it right initially, forming correct vision-based predictions in their intermediate layers, before changing their minds and favoring text in the final output. We call this "late-layer textual override". The visual information is encoded, it simply does not survive to the output. More intriguingly, we find that how predictions change reveals whether they're correct: 85% of failures shift toward text, while 89% of successes shift toward vision. This directional signature enables a simple but powerful intervention: when we detect a confident visual prediction being suppressed, we restore it. We propose CALRD (Conflict-Aware Layer Reference Decoding), a training-free method that recovers overridden predictions at inference time. Experiments across five MLLMs of varying architectures demonstrate up to 9.4% absolute improvements on conflict benchmarks while largely preserving standard performance, without training or external knowledge. It recovers what the model already knew but failed to preserve.
The rapid advancement of generative AI has made audio deepfakes increasingly indistinguishable from authentic human vocals, posing significant threats to persons-of-interest (POI) such as public figures. Current detection systems primarily rely on generic, black-box models that fail to capture speaker-specific idiosyncratic traits and lack interpretability. In this paper, we propose Phoneme-based Voice Profiling (PVP), a novel personalized defense framework. By shifting the detection paradigm from macro-utterance analysis to micro-phonetic modeling, PVP captures the unique acoustic distributions underlying a POI’s habitual articulatory patterns. Specifically, our framework models speaker-specific phonetic realizations using lightweight Gaussian Mixture Models (GMMs) estimated solely from bona fide reference speech. This design enables data-efficient profiling and robust generalization to previously unseen spoofing attacks without requiring heavy spoof-specific training. Furthermore, we introduce the first large-scale Chinese POI deepfake dataset to benchmark speaker-specific detection. Experimental results demonstrate that PVP significantly outperforms state-of-the-art generic detectors in POI spoofing scenarios, achieving substantial EER reductions while providing fine-grained, phoneme-level interpretability for forensic analysis. Code and data are available at: https://github.com/JunXue-tech/PVP
Large Language Models (LLMs) remain highly vulnerable to jailbreak attacks, yet existing evaluations rely primarily on outcome-level metrics such as Attack Success Rate (ASR), providing limited insight into how and why safety failures occur. We propose an explanation-aware safety framework that augments binary harmfulness detection with structured, human-interpretable explanations capturing severity, strategies, trigger spans, rationales, and derived safety factors. To enable scalable and consistent supervision, we introduce a human--LLM hybrid annotation and canonicalization pipeline. We then fine-tune a compact model to generate canonical explanations alongside harmfulness decisions. Across both seen and unseen benchmark settings, our method improves robustness and explanation fidelity. In jailbreak defense evaluation, our approach reduces ASR to 0.44% on Vicuna-7B and 1.30% on GPT-3.5, outperforming existing defense baselines while also achieving the lowest StrongREJECT scores. Beyond outcome-level gains, the model more accurately recovers diagnostic attributes (e.g., attack strategy, trigger spans, and safety factors) than strong general-purpose LLM baselines. Overall, explanation-aware learning exposes diagnostic dimensions that ASR alone cannot capture and provides a more faithful and actionable foundation for robust LLM safety assessment.
Deep convolutional neural networks achieve remarkable performance by exhaustively processing dense spatial feature maps, yet this brute-force strategy introduces significant computational redundancy and encourages reliance on spurious background correlations. As a result, modern vision models remain brittle and difficult to interpret. We propose Energy-Regularized Spatial Masking (ERSM), a novel framework that reformulates feature selection as a differentiable energy minimization problem. By embedding a lightweight Energy-Mask Layer inside standard convolutional backbones, each visual token is assigned a scalar energy composed of two competing forces: an intrinsic Unary importance cost and a Pairwise spatial coherence penalty. Unlike prior pruning methods that enforce rigid sparsity budgets or rely on heuristic importance scores, ERSM allows the network to autonomously discover an optimal information-density equilibrium tailored to each input. We validate ERSM on convolutional architectures and demonstrate that it produces emergent sparsity, improved robustness to structured occlusion, and highly interpretable spatial masks, while preserving classification accuracy. Furthermore, we show that the learned energy ranking significantly outperforms magnitude-based pruning in deletion-based robustness tests, revealing ERSM as an intrinsic denoising mechanism that isolates semantic object regions without pixel-level supervision. Code is available at https://github.com/Tom-Dvk/ERSM.
Partial Dependence Plots (PDPs) visualize how changes in a single feature affect the average model prediction. They are widely used in practice to interpret decision tree ensembles and other machine learning models. Joint-PDPs extend this idea to pairs of features, revealing their combined effect. Partial Dependence Interaction Values (PDIVs) measure feature interactions. The Any-Order-PDIVs task computes these interactions for every feature subset across all rows of the dataset. We introduce WOODELF++, a unified and efficient approach for computing all these useful explainability tools on decision tree ensembles, building on WOODELF, an algorithm for efficient SHAP computation. By deriving suitable metrics over pseudo-Boolean functions, WOODELF++ can compute PDPs (exact and approximate), Joint-PDPs, and Any-Order-PDIVs in a unified framework. Our method delivers substantial complexity improvements over the state of the art, including an exponential gain for Any-Order-PDIVs. Additionally, we introduce and efficiently compute Full PDPs, which leverage the model’s split thresholds to faithfully capture its behavior across all possible feature values. WOODELF++ is implemented in pure Python and supports GPU acceleration. On a dataset with 400,000 rows, WOODELF++ computes PDP and Joint-PDP up to 6x faster than the state of the art and up to five orders of magnitude faster than scikit-learn. For Any-Order-PDIVs, the gap is even larger: WOODELF++ computes all interaction values in 5 minutes, while the state of the art is estimated to require over 1,000,000 years.
Survival models on networks describe the time-evolving dynamics of complex systems, but existing frameworks suffer from two key limitations: First, many past approaches implicitly assume independent censoring, which can introduce substantial bias. Second, point-process-based models often emphasize only sending behavior and overlook shared latent structure between interacting nodes. Recent work has incorporated copulas into deep survival models to mitigate censoring bias, but these methods do not capture topological role structure. Another line of research combines survival analysis with the Mixed-Membership Stochastic Blockmodel (MMSB) to model heterogeneity via role-specific hazard rates; however, it typically assumes structural independence and therefore misses important correlations between roles. To bridge these gaps, we apply an improved Copula–MMSB mechanism and propose the first Survival Fully-Collapsed Copula Mixed Membership Blockmodel. Our framework jointly models time-to-event dynamics and complex dependencies among latent roles. On the Enron benchmark, it surpasses state-of-the-art survival blockmodels in both predictive likelihood and structural interpretability. Additionally, the model admits an exact closed-form discrete distribution over role pairs, which enables a fully collapsed Gibbs sampler and avoids the slow mixing and poor convergence common in non-conjugate inference. In simulations, this yields a 100× improvement in sampling efficiency over semi-collapsed baselines.
The Segment Anything Model (SAM) provides a new paradigm for visual perception tasks through large-scale pre-training and interactive prompts. However, the quality of mask generation is limited by the difficulty of aligning the optimization direction between the backbone and the prompt, which is especially significant in non-natural domain images with low contrast and indistinct semantic edge structure features, especially in SAR images. Therefore, this paper proposes a coupling optimization mechanism that integrates the originally independent prompt generation process with the backbone into a closed loop through the Point Impact Decomposition (PID) module to guide the iterative optimization of the prompt. Specifically, this paper introduces the concept of PID in the decoder to quantify the contribution of prompt points to the masks, constructing an interpretable reward variable to drive the reinforcement learning prompt optimizer. This optimizer designs a reward function based on hypothesis testing ideas, achieving iterative updates of prompt points, and providing feedback on the effects of mask generation based on prompt points. Experiments show that this mechanism can provide a prompt that fits image features; its performance exceeds the existing baseline, which effectively improves the generalization ability of SAM on SAR images.Code is available at https://github.com/Fm336/PID-SAM.
Discovering shapelets -- i.e., discriminative temporal patterns within time series -- has been widely studied to address the inherent complexity of time-series classification (TSC) and to make model decision-making processes more transparent. However, existing methods primarily focus on population-level shapelets optimized across the entire dataset, which leads to two fundamental limitations: (i) population-level patterns often misalign with instance-specific features, resulting in suboptimal performance and potentially misleading interpretations, and (ii) most methods treat shapelets as independent entities, overlooking important temporal dependencies and interactions among multiple patterns. To address these limitations, we propose INSHAPE, an interpretable TSC framework that discovers variable-length, discriminative temporal patterns specific to each time series. INSHAPE identifies these patterns as non-overlapping segments and models their temporal dependencies, thereby providing clear instance-level interpretations while achieving strong predictive performance. Furthermore, INSHAPE bridges local and global interpretability through a bottom-up approach, aggregating instance-level shapelets into prototypical (population-level) shapelets. Extensive experiments on 128 UCR and 30 UEA benchmark datasets show that INSHAPE consistently outperforms state-of-the-art shapelet-based methods while providing more intuitive and interpretable insights.
Many state‑of‑the‑art anomaly detection models operate as black boxes, limiting interpretability and hindering reliable deployment. While recent advances in explainable artificial intelligence have focused on explaining why individual instances are detected as anomalous, comparatively little attention has been paid to how such explanations can be systematically exploited to improve the detectors themselves. To address this gap, we propose EAD‑CE (Enhancing Anomaly Detection with Counterfactual Explanations), a model‑agnostic, closed‑loop framework that tightly integrates detection, explanation, and improvement. Specifically, given a trained anomaly detector and its detected anomalies, EAD‑CE generates minimal and semantically meaningful counterfactual explanations that reveal how targeted feature perturbations influence anomaly scores or decisions. Feature importance inferred from these counterfactual explanations is then used to guide a dynamic feature-weight optimization process, enabling detector refinement without modifying its underlying architecture. Extensive experiments on nine real‑world datasets and three anomaly detection models demonstrate that EAD‑CE accurately identifies anomaly‑driving features, substantially enhances interpretability, and consistently improves detection performance (average AUC increases by 6.9%, up to 23%). Implementation details are provided to support reproducibility.
We propose a modular logical framework for both reasoning with and measuring inconsistency in propositional knowledge bases. The framework extends propositional logic with markers attached to occurrences of atoms and interpreted as pointers to worlds. This allows different occurrences of the same atom to be evaluated in different contexts unless they share a marker. We define paraconsistent entailment relations via abnormality functions that associate each model with a set of deviations from a preferred behavior. Each entailment relation is then defined with respect to models that are minimal under set inclusion. We show that suitable markings and abnormality functions capture several existing forms of inconsistency-tolerant reasoning, including entailment based on maximal satisfiable subsets and the minimally inconsistent Logic of Paradox. We also show how a range of inconsistency measures can be expressed in the same setting. Thus our framework provides a uniform basis for diverse approaches to inconsistency handling and measurement.
Understanding when fair allocation mechanisms lead to stable coalition structures is fundamental for designing robust multi-agent systems. We investigate the compatibility between stable allocations in transferable utility (TU) games and the stability of coalition structures in induced hedonic games, where agents' preferences over coalitions are derived from payoffs assigned by a fixed allocation rule applied to all subgames of the TU game. We analyze FX-FE strong Nash stability (SNS) in induced hedonic games, which captures a strong form of robustness under free-exit and free-entry conditions. We show that any efficient allocation rule ensuring core membership for the grand coalition induces a hedonic game that guarantees the existence of an FX-FE strong Nash stable partition. Examining the Shapley value, we further show that when it lies outside the core, the resulting hedonic game may or may not possess FX-FE strongly Nash stable partitions, highlighting a sensitive interaction between Shapley-based fairness and coalition-level stability. Our framework bridges SHAP methodology from explainable AI with hedonic coalition theory, providing theoretical foundations for understanding when fair allocation mechanisms shape coalition-level incentives in a way that ensures strategic stability in team formation and multi-agent systems.
Instance-wise feature selection (IWFS) identifies informative features for each instance, improving generalization by discarding irrelevant information and enhancing interpretability through personalized explanations. Most IWFS methods adopt a selector--predictor architecture, where a selector generates instance-specific masks to guide prediction. This often leads to co-adaptation, in which the selector encodes label information into the mask, resulting in spurious correlations and unfaithful explanations. Existing methods also struggle to capture diverse local patterns, which is critical for IWFS under heterogeneous sparsity. We propose VIBMask, a unified IWFS framework by the variational information bottleneck. VIBMask mitigates co-adaptation by penalizing mutual information between unselected features and the label, and improves expressivity via an ensemble of diverse selectors that capture heterogeneous sparse patterns. We further derive a novel variational lower bound for discrete masks, enabling efficient end-to-end training through reparameterization. Experiments on synthetic and real datasets show that VIBMask consistently outperforms state-of-the-art IWFS methods in both predictive accuracy and informative feature discovery.
Detecting text generated by large language models~(LLMs) is crucial but challenging. Existing detectors depend on impractical assumptions, such as white-box settings, or solely rely on text-level features, leading to imprecise detection ability. In this paper, we propose a simple but effective and practical LLM-generated text detection method, VaryBalance. The core of VaryBalance is that, compared to LLM-generated texts, there is a greater difference between human texts and their rewritten version via LLMs. Leveraging this observation, VaryBalance quantifies this through mean standard deviation and distinguishes human texts and LLM-generated texts. Comprehensive experiments demonstrated that VaryBalance outperforms the state-of-the-art detectors, i.e., Binoculars, by up to 34.3% in terms of AUROC, and maintains robustness against multiple generating models and languages.
Tool-augmented methods aim to enhance the reasoning capabilities of large language models (LLMs) by invoking external tools, which can be broadly categorized into training-free and training-based methods. Training-free methods can directly instruct LLMs to invoke external tools, but they exhibit limited generalization in dynamic environments. In contrast, training-based methods use a teacher model to generate tool-use trajectories and fine-tune a student model to distill the trajectories that can generalize across tasks. However, they still face two major challenges: (1) Knowledge boundary mismatch: teacher trajectories may contain steps that the student cannot execute without tools due to their mismatch of knowledge boundaries. (2) Tool overuse propagation: the student inherits the misaligned tool-use strategy from teacher model. To address these challenges, we propose RADAR, an automated framework for knowledge boundary discovery and tool overuse mitigation. First, RADAR performs semantic anchoring to select representative seeds. Second, simulated annealing probing is employed to actively explore the student model’s tool-use boundary and acquire sparse, verified student-aware labels. Third, it globally propagates these labels and rewrites conflicts, yielding deployment-aligned decisions with fewer unnecessary tool calls. We conduct experiments on multiple benchmarks, compared to SOTA method, RADAR reduces average tool-use by 52.7% while improving accuracy by 3.2% across three models.
Reconstruction-based frameworks are widely adopted in Time Series Anomaly Detection (TSAD), assuming that models reconstruct normal behavior well but yield larger errors on anomalies. However, in unsupervised TSAD, minimizing reconstruction loss alone often breaks this assumption. Models tend to overfit local patterns for trivial reconstruction and fail to capture the global dependencies that characterize normal behavior. Consequently, anomalies that violate global dependencies can also be reconstructed well, leading to a misalignment between reconstruction and detection. To address this challenge, we propose AnoMamba, a novel TSAD framework that aligns reconstruction with anomaly detection by enhancing global dependency modeling. Specifically, AnoMamba employs patch embedding to reduce local redundancy and introduces a Mamba variant with global step-size reweighting to select meaningful global dependencies. The reweighting process is guided by multi-scale long-tail priors, which adaptively balance global and local dependencies to mitigate overfitting in the unsupervised setting. Extensive experiments on both univariate and multivariate benchmarks demonstrate that AnoMamba consistently outperforms state-of-the-art methods in both accuracy and efficiency, while offering interpretability through its hidden attention map.
Predicting cross-sectional stock returns is challenging due to low signal-to-noise ratios and evolving market regimes. Classical factor models offer interpretability but limited flexibility, while deep learning models achieve strong performance yet often underutilize financial priors. We address this gap with PRISM-VQ (PRior-Informed Stock Model with Vector Quantization), a dynamic factor framework that integrates expert prior factors, vector-quantized discrete latent factors learned from cross-sectional structure, and a structure-conditioned Mixture-of-Experts to generate time-varying factor loadings. Vector quantization acts as an information bottleneck that suppresses noise while capturing robust market structure, with discrete codes serving both as latent factors and as routing signals for temporal expert specialization. Experiments on CSI 300 and S&P 500 show consistent improvements in cross-sectional return prediction and portfolio performance over strong baselines while preserving interpretability. Our code is available at https://github.com/finxlab/PRISM-VQ.