CompanionBench: A Theory-Anchored, Real-World-Grounded Benchmark for AI Emotional Companionship

LLM companions are deployed at scale in personally consequential settings, yet poorly evaluated. Existing benchmarks use hand-authored scenarios and prompted simulators, aggregate empathy into one score, and overlook judge biases such as same-family favoritism and scale drift. We introduce CompanionBench, an interactive bilingual benchmark. To our knowledge, it is the first companion benchmark to ground both its scenarios and a trained user simulator in de-identified real-world data. A hidden disclosure gate branches each persona's trajectory on the agent's own behavior, controlling the interaction state space without scripting dialogue. We operationalize ten capabilities derived from 25 theories across psychology and counseling, four of them not graded explicitly by prior work: holding ambiguity, selfobject responsiveness, positive resonance and calibrated challenge. Agents are assessed on two complementary axes: a subjective ten-capability rubric and a deterministic measure of whether deeper disclosure was earned. A cross-family panel dilutes same-family favoritism; an Item Response Theory model separates agent quality from judge severity. Theory fixes what to measure and how personas are structured; real data supply events, history, and profiles -- coverage from theory, authenticity from data. Rankings are reproducible in both languages (rho = 0.996 ZH / 0.953 EN). Evaluating 28 agents reveals capability-level differences obscured by aggregate scores. Emotion regulation and calibrated challenge remain common weaknesses; holding ambiguity discriminates most. Role-play agents rank near the bottom: immersion does not imply relational competence. Across agents, the dominant failure mode is substituting surface warmth for substantive relational support. We will release 500 Chinese-English parallel pairs and the evaluation code.

Fenced Citation-Context Retrieval for Case Law: Temporal Leakage and Degree Control Across Two Jurisdictions

Prior case retrieval (PCR) aims to identify the precedent cases relevant to the facts of a query case. Incoming citation context, the text with which later cases characterize a case when citing it, is a powerful relevance signal, yet it is typically evaluated without a temporal constraint, so the retriever is credited with citations made after the query. We introduce a zero-training, temporally fenced retriever that augments BM25 with incoming citation context restricted to citations predating the query, together with a temporal-admission decomposition that quantifies the phantom fraction: the share of a citation-context gain attributable to citations not known to predate the query. Experiments span two jurisdictions, U.S. federal (CLERC) and European (ECtHR-PCR) case law. On ECtHR-PCR, without any training, the fenced retriever outperforms a strong degree-controlled baseline across the full recall ladder, and a temporal-admission decomposition attributes 14.9% (validation) of an unfenced citation-context gain over BM25 to citations not known to predate the query. Citation-context retrieval must therefore be temporally fenced and degree-controlled before its reported gains can be interpreted.

ExpoMotion: A Large-Scale Benchmark and A Householder Projection Network for Multi-Exposure Fusion

Multi-Exposure Fusion (MEF) effectively extends dynamic range, but practical deployment is hindered by motion-induced ghosting and the scarcity of high-quality dynamic benchmarks. Current benchmarks largely neglect dynamic scenes and lack reliable ground truth, making it difficult to handle the complexity of real-world motions. In response, we introduce ExpoMotion, a large-scale benchmark designed to evaluate deghosting capabilities. Comprising 1,738 sequences and 10,909 images across diverse environments, it covers a wide range of motions and provides high-fidelity GTs constructed through an expert-guided acquisition pipeline. To tackle the complex dynamics and extreme conditions captured in this benchmark, we propose the Householder Orthogonal Projection network (HOP), which revisits MEF deghosting from a mathematical perspective via Householder transformation, decoupling multi-frame alignment into exposure pre-alignment and ghost filtering. Specifically, the Global Priors Illumination Alignment (GPIA) module first rectifies drastic dynamic range discrepancies by utilizing global statistics for exposure harmonization. Regarding ghost removal, our Householder Orthogonal Attention (HOA) models artifacts as orthogonal perturbations. By employing a dynamic Householder reflector, HOA effectively projects ghosts out of the feature manifold while preserving high-frequency details. Experiments demonstrate that our ExpoMotion dataset enables superior generalization and artifact-free detail restoration, while also validating the effectiveness and efficiency of the HOP method. The dataset and code are available at https://github.com/Leo-LiuYao/ExpoMotion.

There and Back Again: A Flexible-Frame Transformer for Multi-Exposure Fusion

Multi-exposure fusion (MEF) brings the dynamic range of conventional cameras closer to that of human vision, producing images with rich scene content. Given the large variability in scene luminance, exposure strategies often require different numbers of frames to capture the full radiance range faithfully. However, conventional MEF techniques are typically designed for a fixed number of inputs, forcing deployment systems to maintain separate models for different frame-count requirements, which undermines deployment efficiency. To address this limitation, we propose FreeMEF, the first flexible-frame transformer for MEF that seamlessly accommodates varying numbers of input exposures without retraining or architectural changes. The proposed approach consists of two key modules. First, we introduce a recurrent state space module (RSSM) that sequentially fuses features from arbitrary sequences via adaptive alignment and state-space recurrent modeling, thereby providing global information guidance for the subsequent restoration. Second, we devise a global feature guided block (GFGB) incorporating an extremity-aware hybrid attention (EAHA) and an affine-injection feed-forward network (AFFN), which effectively resolves the similarity paradox while simultaneously optimizing contrast and brightness regulation. Extensive experiments on three benchmark datasets demonstrate the effectiveness of our method, which performs favorably against state-of-the-art methods both quantitatively and qualitatively.

Pose-Agnostic Robotic Functional Grasping via Observation-Action Canonicalization

Functional robotic grasping requires a policy that generalizes across diverse object geometries and poses while maintaining task-specific contact precision. We study this challenge through mug-handle grasping, where thin handles, instance variation, and upright or inverted placements make both perception and control sensitive to object configuration. Grasp pose detection methods operate open-loop and are sensitive to estimation errors on thin handle structures. Learned visuomotor policies must implicitly learn to handle the coupled variation in visual appearance and action direction induced by different object placements, limiting generalization. We propose AnyMug, a canonicalized visuomotor reinforcement learning framework for functional grasping that trains a single closed-loop policy entirely in simulation and deploys it zero-shot on a real robot. AnyMug introduces observation-action canonicalization, which transforms both the depth observation and the predicted end-effector action into a shared object-centric frame. The policy therefore sees a consistent mug-centered view and emits actions in a canonical direction regardless of mug placement, allowing the same grasping behavior to be reused across configurations. A handle-aware reward further encourages precise approach, gripper alignment, and opposing-finger placement, while a pose curriculum and domain randomization improve training stability and sim-to-real transfer. In simulation, AnyMug achieves over 93% success rate on both unseen upright and inverted mugs and transfers zero-shot to a real Franka Panda, reaching 80% success rate on 5 held-out physical mugs across both pose categories.

HydraHead: From Head-Level Functional Heterogeneity to Specialized Attention Hybridization

The quadratic complexity of attention poses a critical bottleneck for long-context processing, spurring interest in hybrid attention designs. Most open-source hybrid models adopt a layer-wise strategy. Yet, prior work has noted the inherent difficulty of integrating Linear Attention (LA) with Full Attention (FA), suggesting that the design space of attention hybridization remains underexplored. To probe this space, we conduct interpretability analysis and observe that layers exhibit block-wise functional similarity, while individual heads within the same layer display distinct functional specialization despite sharing input features. This head-level heterogeneity suggests that the head dimension provides a natural and principled granularity for fusing heterogeneous attention signals. Building on this insight, we introduce HydraHead, a novel architecture that hybridizes FA and LA along the head axis. HydraHead features two key innovations: (1) an interpretability-driven selection strategy that identifies retrieval-critical heads and preserves FA only for them, and (2) a scale-normalized fusion module that reconciles the distributional gap between FA and LA head outputs. By leveraging a three-stage transfer pipeline with parameter reuse and distillation, we achieve high-performance hybrid models with minimal training overhead. Under a unified training setup, HydraHead outperforms other hybrid designs in long-context tasks while maintaining strong general reasoning. With interpretability-driven head selection, it matches a 3:1 layer-wise hybrid's long-context performance at a 7:1 LA-to-FA ratio. Crucially, trained on only 15B tokens, HydraHead achieves over 69% improvement over the baseline at 512K context length, approaching Qwen3.5, a leading model of comparable size with a native context length of 256K. This highlights the significant scaling potential of head-level hybridization.

Charge as a Construct-Validity Factor in Chinese Legal Case Retrieval: A Cross-Benchmark Audit

Chinese Legal Case Retrieval (LCR) benchmarks grade a reference judgment relevant when its legal characterization matches the query, and strong systems now reach NDCG@10 of 0.85-0.88. Most of the BM25-to-best-trained gap is recoverable with no retrieval model: ranking candidates only by shared primary charge, broken by BM25, closes 99.2% of it on LeCaRDv2 -- with no detectable difference from the best-trained system. This reflects benchmark design: LeCaRDv2 defines top relevance via the crime's key constitutive elements, which encode the charge, so same-charge cases are relevant by construction (relevance lift 4.49; charge-to-relevance macro-AUC 0.871). Holding charge fixed, the trained reranker's advantage over BM25 collapses to a small within-charge residual (+0.026 NDCG@10, cluster-bootstrap CI excluding zero, about a quarter), the only non-definitional positive. The effect is not uniform: the same rule recovers 84.3% on LeCaRDv1 and is out of spec on CAIL2022, with the charge-to-relevance signal weakening in step (macro-AUC 0.871/0.759/0.728); a predicted-charge cascade reproduces 76.6% on LeCaRDv2 but does not transfer. The construct is also cashable at first stage: an exploratory zero-training charge-pool channel lifts LeCaRDv2 recall (R@100 +0.025, wrong-charge controls hurt), reported as a positive control for the confound, not a retrieval method or novelty claim. Charge is thus a high-leverage construct-validity factor at the benchmark level -- not auniform explanation of NDCG@10, and not evidence that any system relies on charge. We package established construct-validity and partial-input checks as a reusable charge-controlled protocol (CCE); on all three benchmarks its triggers come back null or descriptive, behaving as designed. We release the scripts, schema, and protocol so future benchmarks can be screened before their NDCG@10 is read as legal-reasoning ability.

GLIDE: Graph-guided Leap Inference for Diffusion Estimation of Spatio-Temporal Point Processes

Spatio-temporal point processes (STPPs) provide a principled framework for modeling asynchronous events in continuous time and space. Recent diffusion-based approaches offer a flexible alternative to deterministic prediction by modeling complex conditional distributions, but their application to STPPs remains challenging: reverse sampling from pure noise is costly, and weak structural constraints in sparse spatial domains can lead to poorly localized probability mass. We propose \textbf{GLIDE} (Graph-guided Leap Inference for Diffusion Estimation), a conditional diffusion framework for next-event modeling in STPPs. GLIDE organizes historical events into a multi-scale historical graph and encodes temporal evolution and spatial topology through a dual-stream architecture, yielding a structured conditioning context for a dual-branch diffusion denoiser. It further introduces a prior-guided leap inference mechanism, in which a lightweight mean predictor provides a deterministic anchor and the reverse process starts from an intermediate diffusion step instead of from pure Gaussian noise. Experiments on multiple real-world datasets show that GLIDE improves both distribution fitting and next-event prediction, with the largest gains appearing on the spatial side. The results also indicate that prior-guided leap inference substantially reduces reverse-sampling cost while preserving the stochastic generation capability of diffusion models.

FAiT: Frequency-Aware Inverted Transformer for Multivariate Time Series Forecasting

While Transformer-based architectures have established themselves as a dominant paradigm in Multivariate Time Series Forecasting (MTSF), their core self-attention mechanism inherently functions as a low-pass filter, systematically smoothing out high-frequency signals vital for sharp local changes. Recent advancements have increasingly incorporated frequency-domain operations to address this bias, however, most existing designs rely on fixed spectral bases and apply sequence-wise (uniform) modulation, implicitly assuming a time-invariant frequency response. This overlooks a key property of real-world series that their spectral characteristics often evolve over time, making uniform modulation insufficient for capturing fine-grained temporal dynamics. To tackle these limitations, we propose FAiT, a Frequency-Aware inverted Transformer. Specifically, FAiT rectifies the spectral bias internally through Inverted Attention, which interprets the attention map as a learnable low-pass operator and constructs a dedicated complementary high-pass branch by inverting the attention matrix to recover attenuated transient signals. Furthermore, FAiT introduces Dynamic Temporal-Frequency Modulation (DTFM), which synthesizes instance-conditioned weights to adaptively re-calibrate the energy of spectral sub-bands, enabling fine-grained control over evolving multi-scale patterns. Extensive experiments on widely used benchmarks demonstrate that FAiT consistently outperforms state-of-the-art Transformer-based and frequency-enhanced baselines, while maintaining computational efficiency.

Let the Results Speak: A Replication-First Paradigm for LLM Behavioral Benchmarking

Subjective evaluation of LLM behavior -- empathy, restraint, calibrated emotional tone -- is hard. Human inter-rater agreement on such qualities saturates near rho ~ 0.45, and an LLM-as-judge proxy alone risks circularity: a judge sharing the target's training cohort cannot independently verify it. Anchoring validity to a single human-rater consensus does not extend to capabilities where humans themselves disagree. We propose a replication-first paradigm: instead of anchoring on one rater group, we certify the instrument via four orthogonal properties -- reliability across K runs, cross-instrument replication across architecturally distinct judges, historical-footprint calibration via judges from earlier training cohorts, and pre-registered prediction. We test it on emotional accompaniment by letting the rubric self-evolve data-driven across iterations: the dimensions are not pre-stipulated and the procedure stabilizes to a 9-dimension set. Pre-registration applies to 10 falsifiable hypotheses and 11 forward predictions, committed before any test data was collected. Applied to 49 models across 8 families, the paradigm surfaces what aggregate scores hide. On advice-restraint -- whether a model refrains from giving unsolicited solutions in empathic contexts -- gpt-5 falls 1.87 points from gpt-4.1 and Opus-4.7 falls 0.629 from Opus-4.6, while aggregate scores stay flat. The regression survives three user-proxy swaps (95% of magnitude), replicates across a 5-family judge stack and a 17-month cohort gap, and persists on 74 held-out real ESConv conversations (rho in [0.749, 0.850]); the instrument reaches ordinal Krippendorff alpha = 0.91. As a by-product, the paradigm acts as a saturation-source diagnostic, separating instrumental ceilings (breakable by rubric refinement) from structural ceilings (needing scenario or roster intervention).