In this paper, we propose Concentrate and Concentrate (CaC), a coarse-to-fine anomaly reward model based on Vision-Language Models. During inference, it first conducts a global temporal scan to anchor anomalous time windows, then performs fine-grained spatial grounding within the localized interval, and finally derives robust judgments via structured spatiotemporal Chain-of-Thought reasoning. To equip the model with these capabilities, we construct the first large-scale generated video anomaly dataset with per-frame bounding-box annotations, temporal anomaly windows, and fine-grained attribution labels. Building on this dataset, we design a three-stage progressive training paradigm. The model initially learns spatial and temporal anchoring through single- and multi-frame supervised fine-tuning, and then is optimized by a reinforcement learning strategy based on two-turn Group Relative Policy Optimization (GRPO). Beyond conventional accuracy rewards, we introduce Temporal and Spatial IoU rewards to supervise the intermediate localization process, effectively guiding the model toward more grounded and interpretable spatiotemporal reasoning. Extensive experiments demonstrate that CaC can stably concentrate on subtle anomalies, achieving a 25.7% accuracy improvement on fine-grained anomaly benchmarks and, when used as a reward signal, CaC reduces generated-video anomalies by 11.7% while improving overall video quality.
Jeffrey Qin, Zhuangfei Gao, Wasu Top Piriyakulkij, Mia Radovanovic, Jessica Sommerville, Kevin Ellis, Marta Kryven
Children expertly construct mental models under uncertainty despite limited world knowledge. Which computational principles make this possible? We study this question in an interactive physical Box Task in which participants infer a latent cause through interaction with a stochastic and partially observed environment. The Box Task was originally designed to study children's puzzle-solving qualitatively. We contribute a computational analysis of behavior that formalizes the cognitive process by which children generate new hypotheses and respond to environmental uncertainty as inference unfolds. Our modeling framework is based on Bayesian particle-based inference, which maintains a small set of working hypotheses and continually generates new ones. We give two complementary implementations of hypothesis representations within this framework: (1) as constraints on evidence, and (2) as executable programs evaluated against evidence. Using the constraint-based formulation, we show that the model needs to account both for subjective reliability of evidence and online hypothesis generation to explain key signatures of children's behavior. Using the programmatic formulation, we show that hypotheses may take the form of partial compositional rules to guide information gathering when evidence interpretation is hard. The programmatic hypotheses are implemented using LLM-based program synthesis, evaluated across three prompt variants and several LLM backends, demonstrating robustness of our approach. Together, these results extend current models of children's inference to account for subjective reliability of evidence and open-ended hypothesis spaces, and provide a computational analysis of the form generated hypotheses may take in practice.
Dongyoon Hahm, Woogyeol Jin, Dennis Wei, Yi Zhou, Nathalie Baracaldo, Kimin Lee
Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior.
Environmental, Social, and Governance (ESG) principles are reshaping the foundations of global financial governance, transforming capital allocation architectures, regulatory frameworks, and systemic risk coordination mechanisms. However, as the core medium for assessing corporate ESG performance, the ESG reports present significant challenges for large-scale understanding, due to chaotic reading order from slide-like irregular layouts and implicit hierarchies arising from lengthy, weakly structured content. To address these challenges, we propose Pharos-ESG, a unified framework that transforms ESG reports into structured representations through multimodal parsing, contextual narration, and hierarchical labeling. It integrates a reading-order modeling module based on layout flow, hierarchy-aware segmentation guided by table-of-contents anchors, and a multi-modal aggregation pipeline that contextually transforms visual elements into coherent natural language. The framework further enriches its outputs with ESG, GRI, and sentiment labels, yielding annotations aligned with the analytical demands of financial research. Extensive experiments on annotated benchmarks demonstrate that Pharos-ESG consistently outperforms both dedicated document parsing systems and general-purpose multimodal models. In addition, we release Aurora-ESG, the first large-scale public dataset of ESG reports, spanning Mainland China, Hong Kong, and U.S. markets, featuring unified structured representations of multi-modal content, enriched with fine-grained layout and semantic annotations to better support ESG integration in financial governance and decision-making.
Long-horizon language agents are becoming increasingly important for tasks that require persistent interaction, planning, and reasoning over extended trajectories. External memory plays a central role in these agents by preserving past states, transitions, and experiences beyond the limited context window. Existing memory systems, however, are typically evaluated mainly by task success and token cost, with much less attention to how memory usage is distributed across stored information. We show that repeated retrieval can produce heavy-tailed memory usage, where a small set of states is accessed frequently while many others are rarely retrieved and more prone to prediction errors. We study this phenomenon through a systematic tail audit across retrieval policies and graph structures. We find that memory concentration is reproducible but strongly policy-dependent, ranging from log-normal-compatible patterns under random walks to truncated-power-law-compatible heavy tails under semantic LLM policies. Motivated by these findings, we propose Core-Tail Memory Controller (CTMC), a rank-based memory controller that allocates prompt budget with a single exponent $τ$ while preserving summarized information from the tail. Across Synthetic Graph World and ALFWorld, CTMC substantially reduces memory usage while preserving performance, and on LongMemEval it reduces prompt tokens by 24.48% while maintaining aggregate accuracy. These results show that heavy-tailed memory usage is both a measurable property of long-horizon agent memory and a practical signal for improving memory efficiency.
Zhanzhan Zhao, Yi Yang, Xiao Jia, Zeyun Dong, Chenzhang Wang
As large language models (LLMs) blur the boundary between cognition and computation, they reshape knowledge production and collective life. We combine a curated review of 198 papers with computational mapping of 47,719 formally published papers from five databases. Text embeddings, clustering, and topic modeling organize the curated literature into 13 subcategories across three domains. LLM as Social Minds examines mind-like reasoning and socially grounded behavior in individual models. LLM Societies explores how interactions among agents produce collective behavior and social organization. LLM-Human Interactions examines how people perceive and engage with LLMs, and how interaction reshapes human thinking, creativity, and relationships. Structural topic modeling of the larger corpus traces how research attention shifts over time and differs across publication venues. We present a reproducible framework that identifies the field's underlying structure and enables its continuous refinement as new studies emerge. The review also identifies a gap in current understanding: the mechanisms underlying many of LLMs' observed cognitive and social behaviors remain unclear. Human-like performance can then be mistaken for understanding, obscuring the norms, biases, and structures of power that LLMs reproduce. We call for shared benchmarks and robustness checks across the three domains to build a reliable evidence base for understanding LLMs' social behaviors and informing their design and use.
Liuji Chen, Zeyu Zhang, Xinyuan Zhang, Shuai Nie, Qiang Liu, Shu Wu, Liang Wang
Personalized text generation requires models to capture user-specific writing styles from historical data. Existing approaches based on retrieval, parameter-efficient fine-tuning, or activation steering either introduce inference and storage overhead or struggle to separate stylistic signals from semantic content. We propose GLASS, a training-free framework for personalized generation via Global-Local Activation Steering with Sparse priors. GLASS uses sparse autoencoders to extract a global user-style prior from historical responses and constructs local contrastive style vectors over clustered interaction scenarios. During inference, it jointly injects global and local vectors into different model layers, enabling context-aware personalization without retrieval or parameter updates. Experiments on LaMP and LongLaMP show that GLASS outperforms retrieval-, fine-tuning-, and steering-based baselines across ROUGE metrics and LLM-as-judge evaluations. Further analyses show that SAE-based representations are more robust to topic and length shifts, suggesting better disentanglement of stylistic information from semantic residue.
Long-horizon LLM agents must maintain and update a world model over extended sequences of observations, actions, decisions and tool calls. We study the dynamics of long horizon agents from three perspectives: self-organized criticality, weak chaos, and metastable belief dynamics. Through extensive experiments on controlled puzzles, tool-use tasks, embodied environments, multi-hop retrieval, general-assistant reasoning, and Game of Life, we study how world-model errors develop over long interactions. We find that externally increased task pressure can sharply raise the probability of collapse, while naturally occurring pressure is only a modest predictor of later failure. Errors are temporally correlated rather than independent, and their propagation changes with dependency depth while remaining bounded in finite tasks. We also find that locally valid actions can continue after the agent's overall task state has become incorrect, that error structure depends on task topology and horizon, and that belief states in embodied tasks are shallow rather than strongly trapped in incorrect states. Larger horizons allow larger error cascades, but their scale remains constrained by the task and the model's capability. Overall, we observe power-law behavior, bounded divergence consistent with weak chaos, and metastable belief dynamics across long-horizon agent interactions. We call this perspective world model science: the empirical study of self organized criticality in long-horizon agent world models.
Rubric-based rewards provide fine-grained, interpretable supervision for open-ended reinforcement learning, where holistic scalar rewards struggle to capture multidimensional response quality. Static rubrics can lose discriminative power as the policy distribution evolves, while existing policy--rubric co-evolution methods rely on globally prescribed update schedules that ignore heterogeneous reward informativeness. We introduce RACE, a Reward-Adaptive Co-Evolution framework. At the group level, per-rollout-group training-signal routing classifies groups as informative, diversity-collapsed, or rubric-failed, updating the policy only with informative groups while caching rubric failures. At the phase level, event-triggered phase switching enters rubric optimization upon persistent reward saturation and returns when rubric-reward improvement plateaus. Rubric phases train the generator on cached pairs using a frozen judge and a scale-normalized pairwise-margin reward. Across both levels, online threshold adaptation adjusts routing and switching thresholds through proportional control. Under a unified evaluation protocol, RACE achieves the highest average performance and outperforms the strongest fixed-alternation baseline on all eight tasks, while improving rubric discrimination.
Effective multi-domain adaptation of large language models requires prioritizing relevant domain knowledge while minimizing interference from irrelevant domains. Existing approaches commonly use Low-Rank Adaptation (LoRA) to train a separate adapter for each domain and combine these adapters within one model. However, they either weight adapters equally regardless of the input domain or dynamically weight adapters while still activating all of them including irrelevant ones. Additionally, while efficient, their low-rank structure limits performance. In contrast, upcycling is a model expansion method that adds experts and a router to a dense Transformer layer offering larger capacity than LoRA approaches. In this paper, we propose Domain-Specialized Upcycling (DS-Upcycling) for multi-domain adaptation. We introduce a new domain-aware supervision that allows the router to effectively learn to distinguish domains by selecting the appropriate experts. During fine-tuning on each new domain, we update only the corresponding expert and router while freezing all other parameters. Experiments show that our method achieves the highest average performance across the adapted domains compared to baselines while retaining general performance. With just 256 examples per domain for router supervision, DS-Upcycling achieves 90-98% routing accuracy after adaptation compared with 43-70% for routers trained without our approach.
Closed-loop AI scientists can generate candidate designs at low marginal computational cost, whereas reliable feedback may require wet-lab synthesis, characterization, or high-fidelity computation. Addressing this imbalance through custom laboratory automation remains infrastructure-intensive and costly, while replacing new experiments with a fixed surrogate leaves persistent model errors that can be amplified by optimization. We propose \emph{online surrogate repair} (OSR), a closed-loop algorithm that uses sparse high-fidelity evaluations to update the surrogate throughout a longer agent search conducted primarily with inexpensive surrogate feedback. An acquisition rule selects which designs from the agent's accumulated proposals receive high-fidelity evaluation, and the resulting labels update the surrogate used in subsequent episodes. Across controlled synthetic environments, we demonstrate that improving global surrogate fit does not necessarily reduce maximum regret, whereas Q90-UCB and expected improvement (EI) substantially reduce regret by directing evaluations toward regions that determine the optimizer's decisions. \osradded{Across LLM and Chemeleon workflows, Online EI reduces the amount of oracle queries required by 84.3--90.3\% for matching discovery counts and by 77.9--81.6\% for matching the quality of the best material found.} Online surrogate repair introduces a third feedback regime between fixed-surrogate operation and high-fidelity feedback after every episode, separating the frequency of high-fidelity evaluation from the duration of the agent's search.
LLM agents increasingly operate in modular systems, where planning requires coordinating multiple functional components whose actions can causally change downstream states and behaviors. This makes intervention-aware prediction essential: an agent must anticipate not only what is likely to happen next, but what will happen as a consequence of its own actions. We find that purely observational world models can incur irreducible intervention-time error in such systems, even when they accurately model observational trajectories. To address this limitation, we propose CausalCompose, a causal world-model framework that equips modular LLM agents with causal structure for intervention-aware prediction and planning. We theoretically show that causal composition can overcome this intervention gap with sufficient evidence and controlled local error. Experiments across diverse modular agent environments validate these results: incorporating causal structure consistently improves intervention-sensitive reasoning and planning, with particularly clear gains in structured tool-use systems where causal dependencies directly govern action execution. Our results show when causal world models help modular LLM agents: when causal dependencies are identifiable and actionable for planning.
Robot foundation models increasingly rely on flow-based action generation for general-purpose manipulation, yet the generated actions are not guaranteed to satisfy physical constraints such as joint limits and collision avoidance. This problem is especially challenging when the feasible set cannot be described analytically and is accessible only through black-box feasibility oracles. We study this practical oracle-only setting and propose an oracle-guided differentiable safeguard for safe flow-matching robot policies. The safeguard uses binary feasibility queries to correct unsafe updates in the forward pass, while extracting local boundary information to provide constraint-aware learning signals in the backward pass. Combined with discrete-time flow matching, the same safeguard is used during training and inference, enabling forward--backward aligned policy learning without explicit or differentiable constraint models. Experiments on obstacle-aware robotic manipulation demonstrate 100% constraint satisfaction while preserving task performance.
The way software gets built is changing dramatically. AI agents now create pull requests, select dependencies, refactor codebases, interpret test failures, and execute infrastructure changes across the entire Software Development Life Cycle (SDLC). AI agents are autonomous team members, but unlike human engineers, they are ephemeral - their reasoning, decisions, and tool calls vanish the moment the session ends. Modern Security incidents involving intent mismatch or injection are becoming increasingly common - forcing us to address AI Agents as an insider threat. No widely adopted standard yet enables agentic provenance and governance and its obvious that you can't hold an ephemeral process accountable. The git log was designed for human commits. It shows what changed. It does not show what an AI agent did, which model it used, which tools it called, or whether anyone was accountable for the outcome. Every governance framework, every audit trail, every compliance control in the SDLC was built around one assumption: a human made the decision. That assumption no longer holds. This shift introduces 3 main procedural problems (not addressing scale) - Supply-chain vector - The agentic coder is software, so it is a supply-chain security risk. Policy guardrails - How to ensure the agentic coder is compliant with the policies of the organization. Task-alignment - How to align the agentic coder on its tasks and goals. We introduce Agentic Process Evidence (APE), a signed, tamper-evident, machine-readable attestation of agentic processes, linking the sessions to their outcome (process-aware logging). APE enables Policy-as-Code control over the agentic processes, allowing for human in the loop only when it's needed. We support the in-toto style attestation framework, and the protocol is aligned with the relevant regulations
Understanding the internal routing of visual concepts is crucial for the safe and controllable adaptation of generative models. While concept localization has been widely studied in diffusion models, the emerging paradigm of next-scale visual autoregressive models remains largely unexplored. In this paper, we introduce LoCo, the first model-agnostic framework to localize where and when specific concepts emerge within autoregressive models. Driven by the native coarse-to-fine nature of next-scale generation, our method precisely maps conceptual knowledge across three distinct dimensions: Layer, Scale, and Position. To systematically localize and evaluate concept routing without context bias, we propose LoCoBench, a comprehensive dataset spanning 10 diverse categories. Extensive probing on image and video autoregressive models like Infinity, HunyuanImage-3.0, and InfinityStar shows that the localized positions are both interpretable and causally related to concept emergence. Building on these insights, we apply our localization method to three key applications: concept erasure, model personalization, and adversarial concept injection. Experiments demonstrate that our targeted intervention achieves state-of-the-art performance, substantially reducing computational overhead while preserving benign utility. Overall, our findings offer insights into how conceptual knowledge is routed during autoregressive generation, introducing a practical pathway for more interpretable, efficient, and secure adaptation.
Activation steering changes what a language model generates by editing its hidden states at inference time. Its most common form adds one vector, the difference of class means (DiM), to every hidden state. CHaRS (concept heterogeneity-aware representation steering) uses optimal transport instead: an entropic plan pairs the clusters of the two classes, and each hidden state moves along a mixture of the cluster displacements, weighted toward its nearest clusters. We study this family of clustered optimal-transport maps for toxicity mitigation in Gemma-2-2B and Granite-3.1-2B, and ask what the plan adds to DiM and whether that survives INT8 and NF4 quantization. We prove an identity: whatever the plan, the displacements averaged with the cluster sizes as weights equal DiM, so a clustered map can differ from DiM only for tokens whose routing weights depart from the cluster sizes. In the CHaRS protocol they stay close, and CHaRS is statistically equivalent to Mean-AcT, which adds DiM at every layer, in all eight pre-specified equivalence tests. A variant fitted on individual tokens, which routes each token mainly to its nearest clusters, leaves 37-39% less toxicity than DiM on Gemma and 51-61% less on Granite at steering strength 3 (toxic-bert), where DiM already removes 76-85% of it. Two other toxicity classifiers confirm the gain on Granite, but on Gemma only one does. At equal strength the variant has somewhat higher perplexity, but at matched toxicity it is slightly more fluent than DiM. Yet a nearly constant map from the same clusters also leaves about 30% less toxicity, so the role of treating tokens differently remains open. Under INT8 and NF4 the variant still leaves less toxicity than DiM, and transferring it from full precision to the NF4 model instead of refitting it adds no significant toxicity after correction for multiple tests.
Alexander Stern, Tim Rocktäschel, Acyr Locatelli, Laura Ruis
We report that allocating compute to synthetic pre-pretraining (PPT) never outperforms pure natural-language training in our compute-matched experiments. This holds across four PPT curricula (k-Dyck, neural cellular automata (NCA), declarative reference corpora, and code) and three training regimes: data-constrained, data-unconstrained, and a high token-to-parameter-ratio regime (881M parameters trained on 50B tokens). When natural language is scarce, repeated epochs over a 100M-token corpus beat every synthetic curriculum on natural-language validation loss, with no consistent downstream accuracy advantage from PPT. We find that PPT gains reported in prior work are sensitive to the weight initialization scheme. In an NCA reproduction, we recover a 5.4% perplexity improvement from PPT under the original initialization scheme. An alternative scheme improves both methods, but the no-PPT baseline more, closing the gap between them. However, when PPT receives additional compute rather than a share of the natural-language budget, gains persist at small natural-language budgets even with improved initialization, but disappear by 1B tokens. Finally, models forget what they learn during PPT in subsequent natural-language training; interleaving the same data throughout training prevents forgetting, but does not significantly improve natural-language performance. Our results suggest limited benefits of synthetic pre-pretraining as a distinct training stage and motivate rethinking how such data are used in language model training.
Large language model (LLM) agents increasingly support long-horizon tasks by leveraging external memory and past experience. However, existing agents mainly retrieve records or summaries, lacking the ability to form and refine causal understanding of the world. Inspired by human causal cognition, we introduce CauProphet, an agent that develops and updates causal knowledge within its Causal Cognition. For each task, CauProphet combines relevant knowledge in Causal Cognition with current evidence to construct a Causal Belief, a structured belief capturing causal factors and the effects of possible actions. It uses this belief for decision-making and continuously updates both the belief and Causal Cognition as new observations emerge, enabling causal knowledge that persists beyond individual experiences. Evaluations demonstrate that CauProphet progressively acquires causal knowledge and transfers it to future decision-making. In financial research, CauProphet leverages prior discoveries to improve downstream analyses and outperforms all baselines in mean rank correlation with next-day returns, achieving a 25.4% improvement over a baseline agent. It also improves prediction accuracy from 21.3% to 28.6% in dynamic real-world settings. Further experiments validate its ability to perform reliable causal analysis and action selection across scientific and everyday tasks, demonstrating the potential of causal experience learning for next-generation LLM agents. The project repository is available at https://github.com/OpenCausaLab/CauProphet.
Cone beam computed tomography (CBCT) is widely used for image guided and adaptive radiotherapy, but scatter, noise, reconstruction artifacts, and inaccurate Hounsfield units limit its quantitative use. Recent diffusion approaches for CBCT to CT synthesis remain largely image driven, with limited use of the underlying acquisition physics. We propose EPC-RawDC, a conditional 3D latent diffusion framework that combines volumetric generative modeling with two complementary constraints derived from the same scanner specific CBCT forward model. First, geometry equivariant projection constraint exploits the rotational equivariance of the CBCT forward operator by supervising the synthesized CT through correspondingly transformed reference projections. Second, raw data consistency uses noisy projection measurements of the individual acquisition as scan specific supervision during training. The framework was evaluated on two independent clinical cohorts with distinct scanner geometries and projection sources. For JUST, the physics constraints used measured clinical raw CBCT projections acquired with an Elekta system, whereas for SynthRAD2025 the projection data were simulated from the reference CT using a Varian geometry with Poisson photon statistics. EPC-RawDC achieved an SSIM of 0.97 in both cohorts and improved PSNR over DDPM by 2.66 dB on SynthRAD2025 and 2.10 dB on JUST, while reducing mean MAE by 20.3% and 19.0%, respectively. On SynthRAD2025, projection consistency and equivariant errors were reduced by 55.0% and 41.4% relative to DDPM. These results demonstrate that combining a volumetric diffusion prior with scanner geometry and scan specific projection measurements improves quantitative CBCT to CT synthesis across both simulated and measured projection domains. Code available after publication at https://github.com/ALZAHRAALTALIB/EPC-RawDC.
Peter Dolog, Kim G. Larsen, Marius Mikučionis, Frederik F. K. Rasmussen
For more than three decades, UPPAAL has supported the modelling and analysis of real-time and cyber-physical systems. Yet building a suitable model, formalising the intended properties, choosing analysis techniques and interpreting counterexamples remain iterative tasks that demand substantial expertise. Large language models (LLMs) can offer plausible explanations and even solutions, but they are not necessarily correct. We present UPPAAI, an extension of UPPAAL in which users construct, explain, debug and specify models in dialogue with an LLM. The guiding principle is that the user requests, the LLM proposes, explains and orchestrates, and UPPAAL verifies. UPPAAI exposes the UPPAAL editor, queries and verifier through a Tools API. We report on seven case studies, from explaining to a newcomer to learning a controller, covering timed automata, stochastic hybrid automata and strategy learning. LLMs proved to be extremely helpful in not just generating solutions but also pro-actively suggesting relevant UPPAAL features it was not asked about.
Language models handle text well, and rule engines give conclusions with proofs. Systems that combine the two are usually described by example, and the same coupling turns up under different names in agent frameworks, guardrail toolkits and neuro-symbolic research. This document defines thirteen patterns for coupling a language model with an executable rule language, in the catalogue form introduced for object-oriented design: intent, motivation, applicability, participants, collaborations, consequences, implementation, known uses and related patterns. Four patterns concern what crosses the boundary between model and rules, five who controls whom, three how the model takes part in the reasoning, and one how knowledge is kept over time. Five invariants hold for every combination, and five failure patterns name the couplings that break them. We relate the catalogue to published pattern catalogues and taxonomies of hybrid and neuro-symbolic systems and show where they overlap with it. A decision guide leads from a problem to a pattern. It rests on our measurements, including the cases in which a model alone did as well as the hybrid or better, and it reports what each design cost.
Daniel Strauss, Nikolaos Sotirakis, Zhenshan Bing, Alois Knoll
Memory in reinforcement learning raises two problems. Attention finds the relevant item by scoring the whole store, which is costly for robots under latency and energy limits. When a retrieved memory pays off, credit for acquiring it must cross the whole sequence. We argue that both are governed by one quantity, the value of information, and propose to estimate it from archived rollouts with a learned world model. We obtain that, when read at the observation, it yields a provable lower bound on the optimal value of the acquiring action, which floors its TD target and attributes this floor to the observation that earned it. Read at the decision, it defines the optimal retrieval set. A pilot retrieval system, to be trained and improved with this signal, solves a memory task with reduced context.
Jianghan Zhu, Cong Zhang, Ni Zhang, Yifan Yang, Yew Soon Ong, Zhiguang Cao
Automatic operator design (AOD) has increasingly adopted large language models (LLMs) to generate problem-specific search and variation operators for multi-objective combinatorial optimization (MOCO). Yet most methods evolve the operators while leaving the process that discovers them fixed. This limits adaptation as new Pareto trade-offs and operator interactions change where and how the search should proceed. We introduce HEMO, which makes the AOD harness itself a self-evolving optimization target. HEMO maintains a persistent harness state and updates one module at a time, revising the logic that determines which experience to retrieve, which operators to improve, which parents to select, and how to formulate the orchestration of search steps. Downstream solution quality and runtime guide these updates, while a discounted UCB scheduler directs adaptation toward promising modules as their utility changes. A persistent memory filesystem preserves candidate outcomes and diagnostic feedback across harness states, enabling successive states to build on accumulated optimization experience. Experiments across eight MOCO tasks and two structurally distinct evolutionary backbones demonstrate strong performance against conventional evolutionary, neural, and LLM-based AOD baselines, together with effective transfer to related tasks. HEMO extends automatic operator design beyond evolving individual operators to improving the search process that discovers them.
Michela Taufer, Kin Ng, Ian Lumsden, Aashish Pandey, Rajasi Gore, Angelina Ju, Sukanya Moudgalya, Ewa Deelman 等 9 位作者
AI assistants are increasingly embedded in computational notebooks, scientific workflows, software engineering tools, and learning systems, yet human-AI interactions are often treated as transient prompt-response exchanges rather than durable data products, hindering reproducibility, explainability, replay, behavioral analysis, and adaptive AI support. We argue that these interactions form a new data class requiring formal representation and introduce the Human-AI Interaction Description Language (HAIDL), a formal interaction model with a Backus-Naur Form-inspired concrete syntax. HAIDL encodes each interaction as a structured object capturing computational context, user request, prompt payload, AI response, feedback, problem context, provenance, and model metadata. We integrate this vision into HAIDL-X, a middleware for JupyterLab that instantiates HAIDL by capturing user-bot state, building contextualized prompts, interfacing with AI providers, collecting feedback, and storing interactions in a repository. We add a learning layer over HAIDL that maps interaction objects and notebook execution traces into reproducible measures of user behavior, including engagement, performance, help-seeking, error-oriented use, and learning personas. HAIDL-X provides a foundation for treating human-AI collaboration as structured, queryable, replayable, analyzable data.
João Pedro Gandarela, Thiago Rios, Stefan Menzel, André Freitas
Learning mechanism behaviour requires a representation of how connected parts jointly constrain motion. We study whether exposing reusable constraint subassemblies improves prediction from limited data. Assur-KB represents planar mechanisms through irreducible modules, boundary interfaces, and solve dependencies while retaining source geometry. Our theoretical results characterize transfer through module interfaces and establish physical limits arising from missing interface information and accumulated sensitivity. To test the representation claim directly, the experimental design compares matched mechanism-curve pairs across training budgets within the same multilayer-perceptron family. This controlled comparison establishes the utility of the complete representation in the evaluated learning setting. Assur-derived features provide a geometric-mean $1.8\times$ advantage in sample efficiency over flattened adjacency and position inputs across budgets of 50-10,000 mechanisms. Separate experiments establish invariance under tested joint relabelings, higher valid-generation rates under dyadic composition, and a coarser structural index over 10.02 million mechanisms. In summary, the theory and experiments identify how physical constraint structure supports limited-data prediction and which interface conditions govern compositional reuse.
Jorge L. Rodriguez, Victor Angulo Morales, Areej Alwahas, Mariana Elias Lara, Fida Mohammad Thoker, Kasper Johansen, Bernard Ghanem, Fernando T. Maestre 等 9 位作者
Foundation models offer a promising route to transferable remote sensing representations, but many current approaches depend on very large pretraining datasets and fixed sensor configurations, limiting their suitability for ecological and environmental applications, where observations often vary across platforms, spatial and spectral resolutions, and available modalities. We introduce FLORO, a multimodal geospatial foundation model designed to learn transferable representations from a small but highly diverse remote sensing corpus. FLORO is pretrained using masked autoencoding on a heterogeneous combination of Sentinel-1, Sentinel-2, SkySAT imagery, elevation, and UAV-derived data. To accommodate sensor variability, FLORO incorporates availability-aware inputs that indicate which spectral bands and auxiliary modalities are present in each sample, enabling a unified input space across heterogeneous sensor configurations. We evaluated FLORO on the PANGAEA benchmark under a frozen-encoder protocol across scene classification, segmentation, and regression tasks. Despite being pretrained on a smaller corpus than competing foundation models, FLORO achieved strong and stable transfer across optical, optical-SAR, and optical-elevation benchmarks spanning medium-resolution satellite, airborne, and ultra-high-resolution UAV imagery. FLORO obtained the second-best average segmentation performance across six PANGAEA benchmarks, the third-best average rank across three scene-classification benchmarks, and excelled in regression tasks, while qualitative results showed improved preservation of spatial structure in flood, urban, biomass, and canopy-height prediction settings. Overall, these results position FLORO as a practical framework for ecological and environmental monitoring under heterogeneous sensing conditions.
We present deterministic polynomial-space algorithms that solve Max 2-SAT on $n$ variables exactly in time $\Ostar(2^{(1-γ)n})$, for a fixed $γ>0$, on several classes of instances whose clause density may grow with $n$. Our main tool extends the classic DPLL branching algorithm. When the search gives a variable its less promising value, comparing the two values of the variable yields a pruning constraint that is satisfied by some optimal assignment. Accumulating these constraints along each branch prunes all but an exponentially small fraction of the assignments. The classes are formulas with many biased variables, semirandom 2-CNF formulas, and Max-Cut on regular expander graphs and on graphs generated by a stochastic block model. In a semirandom formula, the pairs of variables that may share a clause are arbitrary, and the clauses on these pairs are random. For semirandom formulas and block-model graphs, the algorithms are still exact on every input, and the time bound holds with high probability over the input. To our knowledge, these are the first polynomial-space algorithms with such a running time for nontrivial classes of dense Max 2-SAT instances. Beyond these classes, the pruning constraints can be added to any branching algorithm for Max 2-SAT, and we expect them to be of practical relevance for general formulas, in particular for sparse ones, where each constraint involves only a few variables.
Large language models are increasingly used to support organizational decisions. Despite their rapid and widespread adoption, users often lack a principled basis for assessing whether to rely on a specific recommendation. We introduce \emph{epistemic warrant}, a decision-level construct that characterizes the scope over which a recommendation is supported. We operationalize this construct through a reliance certificate that uses behavioral evidence from four tiers of tests to assess a recommendation's underlying warrant and assign it to a warrant category. We assess this operationalization using complementary forms of construct-validity evidence, including transformation validity, known-groups tests, independent human consensus, discriminant analyses, and robustness across models and alternative implementations. The results show that the certificate behaves in theoretically expected ways: its transformations instantiate their intended relationships, certificate assignments increase across groups prespecified to represent increasing levels of warrant, and stronger assigned warrant is associated with greater independent human consensus. Certificate assignments also capture information beyond verbalized confidence and are not readily explained by decision difficulty. The main conclusions remain robust across models and operationalizations. These findings support the reliance certificate as an operationalization of epistemic warrant for assessing the scope of support for a particular LLM recommendation when ground truth is unavailable.
Heimo Müller, Dominik Steiger, Markus Plass, Andreas Holzinger
Large language models routinely produce factually incorrect or misleading outputs, yet conventional automated benchmarks overlook how users perceive reliability during interaction. We introduce the System Hallucination Scale (SHS), a lightweight instrument modeling the famous System Usability Scale (SUS) and System Causability Scales (SCS). SHS measures perceived reliability across five dimensions: factual accuracy, source reliability, logical coherence, deceptiveness of presentation, and responsiveness to user guidance, using ten balanced Likert-scale items that provide a directional score and consistency diagnostic. In a supervised validation study with 193 participants (386 evaluations across 47 trained experimenters), SHS demonstrated strong internal consistency and high clarity, completing in an average of 4.2 minutes. Participants rated source verifiability and presentation deceptiveness worse than factual accuracy, indicating perceived vulnerabilities lie primarily in traceability and framing. As a proof of concept, SHS assesses human-perceived reliability rather than ground-truth correctness, providing open-source multilingual implementations and an interactive calculator.
LLM-as-a-judge scales evaluation, but reasoning judges are slow and costly. We study JEV-as-a-Judge: evaluation with JEV, a decision-only judge that returns label probabilities instead of text, and whose confidence decides whether to accept its verdict or escalate to a reasoning judge. Against sixteen generative and reward-model judges, with blinded human adjudication, JEV comes within three points of GPT-6 wherever a verdict can be read off the text, at 0.36% of its fee and a 0.15-second median latency, and falls behind where the verdict must be derived, as in math, code, and logic. Its confidence marks this boundary. With a threshold frozen in advance, accepting confident verdicts and escalating the rest is 0.9 points more accurate than GPT-6 on 1,610 held-out pairs at 41% of its fee, and in a pre-specified live test on two new workloads the cascade matches GPT-6's accuracy exactly. Confidence routing weakens on style-adversarial pairs and reference-free prose; we close with a simple recipe for validating thresholds locally.