Large Vision-Language Models (L-VLMs) have demonstrated remarkable performance in various vision and language tasks, including visual question answering (VQA). However, their high computational cost makes them impractical for resource-constrained settings and inference-heavy applications. In contrast, Small Vision-Language Models (S-VLMs) offer efficiency but suffer from a significant performance gap compared to their larger counterparts. In this work, we introduce the Model Parity Aligner (MPA), a novel framework designed to systematically improve S-VLMs by leveraging unlabeled images and effective knowledge transfer from L-VLMs. Instead of traditional knowledge distillation methods that rely on labeled training data, MPA employs a strategic parity-based approach that precisely identifies the knowledge disparities between S-VLMs and L-VLMs, and optimizes training by targeting only these disparities. We conduct extensive experiments on four diverse VQA benchmarks, namely TextVQA, ST-VQA, ChartQA, and OKVQA, each of which requires specialized reasoning capabilities such as text recognition, chart interpretation, and commonsense and factual understanding. Our results demonstrate that MPA consistently enhances the performance of S-VLMs on all benchmarks, reducing the performance gap while maintaining computational efficiency. We make our code publicly available.

The debate around Artificial General Intelligence (AGI) remains open due to two fundamentally different goals: replicating human-like performance versus replicating human-like cognitive processes. We argue that current performance-based definitions are inadequate because they provide no clear, mechanism-focused roadmap for research, and they fail to properly define the qualitative nature of genuine intelligence. Drawing inspiration from the human brain, we propose a new paradigm that shifts the focus from external mimicry to the development of foundational cognitive architectures. We define True Intelligence (TI) as a system characterized by six core components: embodied sensory fusion, core directives, dynamic schemata creation, a highly-interconnected multi-expert architecture, an orchestration layer, and lastly, the unmeasurable quality of Interconnectedness, which we hypothesize results in consciousness and a subjective experience. We propose a practical, five-level taxonomy of AGI based on the number of the first five measurable components a system exhibits. This framework provides a clear path forward with developmental milestones that directly address the challenge of building genuinely intelligent systems. We contend that once a system achieves Level-5 AGI by implementing all five measurable components, the difference between it and TI remains as a purely philosophical debate. For practical purposes - and given theories indicate consciousness is an emergent byproduct of integrated, higher-order cognition - we conclude that a fifth-level AGI is functionally and practically equivalent to TI. This work synthesizes diverse insights from analytical psychology, schema theory, metacognition, modern brain architectures and latest works in AI to provide the first holistic, mechanism-based definition of AGI that offers a clear and actionable path for the research community.

Speculative decoding is a powerful way to accelerate autoregressive large language models (LLMs), but directly porting it to vision-language models (VLMs) faces unique systems constraints: the prefill stage is dominated by visual tokens whose count scales with image resolution and video length, inflating both compute and memory--especially the key-value (KV) cache. We study speculative decoding for VLMs and introduce SpecVLM, a practical system that (1) establishes a strong EAGLE-2-style baseline, EagleVLM, delivering 1.5-2.3 To avoid costly offline distillation corpora, we propose an online-logit distillation protocol that trains the draft model with on-the-fly teacher logits and penultimate features using a combined cross-entropy and Smooth L1 objective, eliminating storage and preprocessing while remaining compute-efficient. This protocol reveals a training-time scaling effect: longer online training monotonically increases the draft model's average accepted length, improving speculative efficiency. Empirically, SpecVLM achieves additional acceleration, culminating in 2.5-2.9 end-to-end speedups within 5 epochs across LLaV A and MMMU, consistently over resolutions and task difficulties, while preserving the target model's output distribution (lossless decoding). Autoregressive decoding underpins many high-quality vision-language models (VLMs) such as LLaV A (Liu et al., 2023), GPT -4 (Achiam et al., 2023), and Gemini (Team et al., 2023), which are widely used for image captioning, visual question answering, and multimodal dialogue. While these teacher models produce high-fidelity outputs, their token-by-token decoding is computationally expensive--an issue that is amplified in multimodal settings because the prefill stage (visual encoding projection token injection) can dominate wall-clock time and memory usage (Li et al., 2025b). Higher image resolutions, denser visual tokenizations, and video inputs dramatically increase the number of visual tokens, which in turn inflates both the prefill cost and the KV cache traffic during decoding.

Successful teamwork depends on interpersonal dynamics, the ways in which individuals coordinate, influence, and adapt to one another over time. Existing measures of interpersonal dynamics, such as CRQA, correlation, Granger causality, and transfer entropy, typically capture only a single dimension: either the synchrony/coordination or the direction of influence between individuals. What is missing is a psychologically meaningful representation that unifies these dimensions and varies systematically with behavior. We propose the "context matrix" as one such representation. The context matrix, modeled within a linear dynamical system, has psychologically interpretable entries specifying how much each individual's current behavior is attributable to their own versus every other group member's past behaviors. Critically, these entries can be distilled into summary features that represent synchrony and directional influence. Evidence for the context matrix as psychologically meaningful is provided in two steps. First, we develop a sequential Bayesian model that infers context matrices from timeseries data and show that it accurately recovers them in noisy simulations. Second, applying the model to human eyetracking data, we demonstrate that summary features of the inferred context matrices capture expected task-based differences in interpersonal dynamics (or lack thereof), predict task accuracy in psychologically reasonable ways, and show some correspondence with existing measures (CRQA and Granger causality). We conclude by situating the context matrix within a broader agenda for modeling interpersonal dynamics in joint action.

Dynamic nonlinear systems exhibit distortions arising from coupled static and dynamic effects. Their intertwined nature poses major challenges for data-driven modeling. This paper presents a theoretical framework grounded in structured decomposition, variance analysis, and task-centric complexity bounds. The framework employs a directional lower bound on interactions between measurable system components, extending orthogonality in inner product spaces to structurally asymmetric settings. This bound supports variance inequalities for decomposed systems. Key behavioral indicators are introduced along with a memory finiteness index. A rigorous power-based condition establishes a measurable link between finite memory in realizable systems and the First Law of Thermodynamics. This offers a more foundational perspective than classical bounds based on the Second Law. Building on this foundation, we formulate a `Behavioral Uncertainty Principle,' demonstrating that static and dynamic distortions cannot be minimized simultaneously. We identify that real-world systems seem to resist complete deterministic decomposition due to entangled static and dynamic effects. We also present two general-purpose theorems linking function variance to mean-squared Lipschitz continuity and learning complexity. This yields a model-agnostic, task-aware complexity metric, showing that lower-variance components are inherently easier to learn. These insights explain the empirical benefits of structured residual learning, including improved generalization, reduced parameter count, and lower training cost, as previously observed in power amplifier linearization experiments. The framework is broadly applicable and offers a scalable, theoretically grounded approach to modeling complex dynamic nonlinear systems.

Abstract--A recent line of research on automated speaking assessment (ASA) has benefited from self-supervised learning (SSL) representations, which capture rich acoustic and linguistic patterns in non-native speech without underlying assumptions of feature curation. However, speech-based SSL models capture acoustic-related traits but overlook linguistic content, while text-based SSL models rely on ASR output and fail to encode prosodic nuances. Moreover, most prior arts treat proficiency levels as nominal classes, ignoring their ordinal structure and non-uniform intervals between proficiency labels. T o address these limitations, we propose an effective ASA approach combining SSL with handcrafted indicator features via a novel modeling paradigm. We further introduce a multi-margin ordinal loss that jointly models both the score ordinality and non-uniform intervals of proficiency labels. Extensive experiments on the TEEMI corpus show that our method consistently outperforms strong baselines and generalizes well to unseen prompts.

Despite their remarkable success in complex tasks propelling widespread adoption, large language-model-based agents still face critical deployment challenges due to prohibitive latency and inference costs. While recent work has explored various methods to accelerate inference, existing approaches suffer from significant limitations: they either fail to preserve performance fidelity, require extensive offline training of router modules, or incur excessive operational costs. Moreover, they provide minimal user control over the tradeoff between acceleration and other performance metrics. To address these gaps, we introduce Dynamic Speculative Planning (DSP), an asynchronous online reinforcement learning framework that provides lossless acceleration with substantially reduced costs without requiring additional pre-deployment preparation. DSP explicitly optimizes a joint objective balancing end-to-end latency against dollar cost, allowing practitioners to adjust a single parameter that steers the system toward faster responses, cheaper operation, or any point along this continuum. Experiments on two standard agent benchmarks demonstrate that DSP achieves comparable efficiency to the fastest lossless acceleration method while reducing total cost by 30% and unnecessary cost up to 60%. Our code and data are available through https://github.com/guanyilin428/Dynamic-Speculative-Planning.

Intelligent Tutoring Systems (ITS), such as Massive Open Online Courses, offer new opportunities for human learning. At the core of such systems, knowledge tracing (KT) predicts students' future performance by analyzing their historical learning activities, enabling an accurate evaluation of students' knowledge states over time. We show that existing KT methods often encounter correlation conflicts when analyzing the relationships between historical learning sequences and future performance. To address such conflicts, we propose to extract so-called Follow-up Performance Trends (FPTs) from historical ITS data and to incorporate them into KT. We propose a method called Forward-Looking Knowledge Tracing (FINER) that combines historical learning sequences with FPTs to enhance student performance prediction accuracy. FINER constructs learning patterns that facilitate the retrieval of FPTs from historical ITS data in linear time; FINER includes a novel similarity-aware attention mechanism that aggregates FPTs based on both frequency and contextual similarity; and FINER offers means of combining FPTs and historical learning sequences to enable more accurate prediction of student future performance. Experiments on six real-world datasets show that FINER can outperform ten state-of-the-art KT methods, increasing accuracy by 8.74% to 84.85%.

Abstract--Long-Horizon (LH) tasks in Human-Scene Interaction (HSI) are complex multi-step tasks that require continuous planning, sequential decision-making, and extended execution across domains to achieve the final goal. However, existing methods heavily rely on skill chaining by concatenating pre-trained subtasks, with environment observations and self-state tightly coupled, lacking the ability to generalize to new combinations of environments and skills, failing to complete various LH tasks across domains. T o solve this problem, this paper presents DET ACH, a cross-domain learning framework for LH tasks via biologically inspired dual-stream disentanglement. Inspired by the brain's "where-what" dual pathway mechanism, DET ACH comprises two core modules: i) an environment learning module for spatial understanding, which captures object functions, spatial relationships, and scene semantics, achieving cross-domain transfer through complete environment-self disentanglement; ii) a skill learning module for task execution, which processes self-state information including joint degrees of freedom and motor patterns, enabling cross-skill transfer through independent motor pattern encoding. We conducted extensive experiments on various LH tasks in HSI scenes. Compared with existing methods, DET ACH can achieve an average subtasks success rate improvement of 23% and average execution efficiency improvement of 29%. More details can be found at: https: //sites.google.com/view/detach-learning. I. INTRODUCTION Long-Horizon (LH) tasks in Human-Scene Interaction (HSI) require continuous planning and cross-domain execution, posing challenges due to their complexity and need for environmental adaptation. These tasks have broad applications in robotics [1], medical intervention [2], and smart homes [2], with canonical examples including dexterous hand manipulation [3] and humanoid whole-body control [4].

Large Language Models (LLMs) have shown remarkable capabilities through two complementary paradigms: Retrieval-Augmented Generation (RAG), which enhances knowledge grounding, and Reinforcement Learning from Verifiable Rewards (RLVR), which optimizes complex reasoning abilities. However, these two capabilities are often developed in isolation, and existing efforts to unify them remain narrow in scope -- typically limited to open-domain QA with fixed retrieval settings and task-specific constraints. This lack of integration constrains generalization and limits the applicability of RAG-RL methods to broader domains. To bridge this gap, we propose UR2 (Unified RAG and Reasoning), a general framework that unifies retrieval and reasoning through reinforcement learning. UR2 introduces two key contributions: a difficulty-aware curriculum training that selectively invokes retrieval only for challenging problems, and a hybrid knowledge access strategy combining domain-specific offline corpora with LLM-generated summaries. These components are designed to enable dynamic coordination between retrieval and reasoning, improving adaptability across a diverse range of tasks. Experiments across open-domain QA, MMLU-Pro, medical, and mathematical reasoning tasks demonstrate that UR$^2$ (built on Qwen-2.5-3/7B and LLaMA-3.1-8B) significantly outperforms existing RAG and RL methods, achieving comparable performance to GPT-4o-mini and GPT-4.1-mini on several benchmarks. We have released all code, models, and data at https://github.com/Tsinghua-dhy/UR2.