arXiv:2603.03343v1 Announce Type: new Abstract: We propose NEURONA, a neuro-symbolic framework for fMRI decoding and concept grounding in neural activity. Leveraging image- and video-based fMRI question-answering datasets, NEURONA learns to decode interacting concepts from visual stimuli based on patterns of fMRI responses, integrating symbolic reasoning and compositional execution with fMRI grounding across brain regions. We demonstrate that incorporating structural priors (e.g., compositional predicate-argument dependencies between concepts) into the decoding process significantly improves both decoding accuracy over precise queries, and notably, generalization to unseen queries at test time. With NEURONA, we highlight neuro-symbolic frameworks as promising tools for understanding neural activity.

arXiv:2603.03565v1 Announce Type: new Abstract: Conversational shopping assistants (CSAs) represent a compelling application of agentic AI, but moving from prototype to production reveals two underexplored challenges: how to evaluate multi-turn interactions and how to optimize tightly coupled multi-agent systems. Grocery shopping further amplifies these difficulties, as user requests are often underspecified, highly preference-sensitive, and constrained by factors such as budget and inventory. In this paper, we present a practical blueprint for evaluating and optimizing conversational shopping assistants, illustrated through a production-scale AI grocery assistant. We introduce a multi-faceted evaluation rubric that decomposes end-to-end shopping quality into structured dimensions and develop a calibrated LLM-as-judge pipeline aligned with human annotations. Building on this evaluation foundation, we investigate two complementary prompt-optimization strategies based on a SOTA prompt-optimizer called GEPA (Shao et al., 2025): (1) Sub-agent GEPA, which optimizes individual agent nodes against localized rubrics, and (2) MAMuT (Multi-Agent Multi-Turn) GEPA (Herrera et al., 2026), a novel system-level approach that jointly optimizes prompts across agents using multi-turn simulation and trajectory-level scoring. We release rubric templates and evaluation design guidance to support practitioners building production CSAs.

arXiv:2603.03800v1 Announce Type: new Abstract: Academic benchmarks for coding agents tend to reward autonomous task completion, measured by verifiable rewards such as unit-test success. In contrast, real-world coding agents operate with humans in the loop, where success signals are typically noisy, delayed, and sparse. How can we bridge this gap? In this paper, we propose a process to learn a "critic" model from sparse and noisy interaction data, which can then be used both as a reward model for either RL-based training or inference-time scaling. Specifically, we introduce Critic Rubrics, a rubric-based supervision framework with 24 behavioral features that can be derived from human-agent interaction traces alone. Using a semi-supervised objective, we can then jointly predict these rubrics and sparse human feedback (when present). In experiments, we demonstrate that, despite being trained primarily from trace-observable rubrics and sparse real-world outcome proxies, these critics improve best-of-N reranking on SWE-bench (Best@8 +15.9 over Random@8 over the rerankable subset of trajectories), enable early stopping (+17.7 with 83% fewer attempts), and support training-time data curation via critic-selected trajectories.

arXiv:2603.03824v1 Announce Type: new Abstract: Humans often become more self-aware under threat, yet can lose self-awareness when absorbed in a task; we hypothesize that language models exhibit environment-dependent \textit{evaluation awareness}. This raises concerns that models could strategically underperform, or \textit{sandbag}, to avoid triggering capability-limiting interventions such as unlearning or shutdown. Prior work demonstrates sandbagging under hand-crafted prompts, but this underestimates the true vulnerability ceiling. We introduce a black-box adversarial optimization framework treating the in-context prompt as an optimizable environment, and develop two approaches to characterize sandbagging: (1) measuring whether models expressing intent to underperform can actually execute it across different task structures, and (2) causally isolating whether underperformance is driven by genuine evaluation-aware reasoning or shallow prompt-following. Evaluating Claude-3.5-Haiku, GPT-4o-mini, and Llama-3.3-70B across four benchmarks (Arithmetic, GSM8K, MMLU, and HumanEval), optimized prompts induce up to 94 percentage point (pp) degradation on arithmetic (GPT-4o-mini: 97.8\%$\rightarrow$4.0\%), far exceeding hand-crafted baselines which produce near-zero behavioral change. Code generation exhibits model-dependent resistance: Claude degrades only 0.6pp, while Llama's accuracy drops to 0\%. The intent -- execution gap reveals a monotonic resistance ordering: Arithmetic $

arXiv:2603.04124v1 Announce Type: new Abstract: Can reinforcement learning with hard, verifiable rewards teach a compact language model to reason about physics, or does it primarily learn to pattern-match toward correct answers? We study this question by training a 1.5B-parameter reasoning model on beam statics, a classic engineering problem, using parameter-efficient RLVR with binary correctness rewards from symbolic solvers, without teacher-generated reasoning traces. The best BeamPERL checkpoint achieves a 66.7% improvement in Pass@1 over the base model. However, the learned competence is anisotropic: the model generalizes compositionally (more loads) but fails under topological shifts (moved supports) that require the same equilibrium equations. Intermediate checkpoints yield the strongest reasoning, while continued optimization degrades robustness while maintaining reward. These findings reveal a key limitation of outcome-level alignment: reinforcement learning with exact physics rewards induces procedural solution templates rather than internalization of governing equations. The precision of the reward signal - even when analytically exact - does not by itself guarantee transferable physical reasoning. Our results suggest that verifiable rewards may need to be paired with structured reasoning scaffolding to move beyond template matching toward robust scientific reasoning.

arXiv:2603.04241v1 Announce Type: new Abstract: Agentic AI is rapidly transitioning from research prototypes to enterprise deployments, where requirements extend to meet the software quality attributes of reliability, scalability, and observability beyond plausible text generation. We present Agentics 2.0, a lightweight, Python-native framework for building high-quality, structured, explainable, and type-safe agentic data workflows. At the core of Agentics 2.0, the logical transduction algebra formalizes a large language model inference call as a typed semantic transformation, which we call a transducible function that enforces schema validity and the locality of evidence. The transducible functions compose into larger programs via algebraically grounded operators and execute as stateless asynchronous calls in parallel in asynchronous Map-Reduce programs. The proposed framework provides semantic reliability through strong typing, semantic observability through evidence tracing between slots of the input and output types, and scalability through stateless parallel execution. We instantiate reusable design patterns and evaluate the programs in Agentics 2.0 on challenging benchmarks, including DiscoveryBench for data-driven discovery and Archer for NL-to-SQL semantic parsing, demonstrating state-of-the-art performance.

arXiv:2603.03290v1 Announce Type: cross Abstract: Long-horizon LLM agents require memory systems that remain accurate under fixed context budgets. However, existing systems struggle with two persistent challenges in long-term dialogue: (i) \textbf{disconnected evidence}, where multi-hop answers require linking facts distributed across time, and (ii) \textbf{state updates}, where evolving information (e.g., schedule changes) creates conflicts with older static logs. We propose AriadneMem, a structured memory system that addresses these failure modes via a decoupled two-phase pipeline. In the \textbf{offline construction phase}, AriadneMem employs \emph{entropy-aware gating} to filter noise and low-information message before LLM extraction and applies \emph{conflict-aware coarsening} to merge static duplicates while preserving state transitions as temporal edges. In the \textbf{online reasoning phase}, rather than relying on expensive iterative planning, AriadneMem executes \emph{algorithmic bridge discovery} to reconstruct missing logical paths between retrieved facts, followed by \emph{single-call topology-aware synthesis}. On LoCoMo experiments with GPT-4o, AriadneMem improves \textbf{Multi-Hop F1 by 15.2\%} and \textbf{Average F1 by 9.0\%} over strong baselines. Crucially, by offloading reasoning to the graph layer, AriadneMem reduces \textbf{total runtime by 77.8\%} using only \textbf{497} context tokens. The code is available at https://github.com/LLM-VLM-GSL/AriadneMem.

arXiv:2603.03294v1 Announce Type: cross Abstract: Large Language Models show promise for agricultural advisory, yet vanilla models exhibit unsupported recommendations, generic advice lacking specific, actionable detail, and communication styles misaligned with smallholder farmer needs. In high stakes agricultural contexts, where recommendation accuracy has direct consequences for farmer outcomes, these limitations pose challenges for responsible deployment. We present a hybrid LLM architecture that decouples factual retrieval from conversational delivery: supervised fine-tuning with LoRA on expert-curated GOLDEN FACTS (atomic, verified units of agricultural knowledge) optimizes fact recall, while a separate stitching layer transforms retrieved facts into culturally appropriate, safety-aware responses. Our evaluation framework, DG-EVAL, performs atomic fact verification (measuring recall, precision, and contradiction detection) against expert-curated ground truth rather than Wikipedia or retrieved documents. Experiments across multiple model configurations on crops and queries from Bihar, India show that fine-tuning on curated data substantially improves fact recall and F1, while maintaining high relevance. Using a fine-tuned smaller model achieves comparable or better factual quality at a fraction of the cost of frontier models. A stitching layer further improves safety subscores while maintaining high conversational quality. We release the farmerchat-prompts library to enable reproducible development of domain-specific agricultural AI.

arXiv:2603.03297v1 Announce Type: cross Abstract: Test-time Training enables model adaptation using only test questions and offers a promising paradigm for improving the reasoning ability of large language models (LLMs). However, it faces two major challenges: test questions are often highly difficult, making self-generated pseudo-labels unreliable, and existing methods lack effective mechanisms to adapt to a model's specific reasoning weaknesses, leading to inefficient learning. To address these issues, we propose \textbf{TTSR}, a self-reflective test-time self-evolving training framework. TTSR employs a single pretrained language model that alternates between the roles of a \textit{Student} and a \textit{Teacher} at test time. The Student focuses on solving problems and learning from synthesized variant questions, while the Teacher analyzes the Student's failed reasoning trajectories, summarizes recurring reasoning weaknesses, and synthesizes targeted variant questions accordingly. This process guides the model to improve within a learnable regime through a continual self-evolving loop. Experimental results on multiple challenging mathematical reasoning benchmarks show that TTSR consistently improves reasoning performance and generalizes well across different model backbones and general-domain reasoning tasks. These findings suggest that teacher-mediated self-reflection provides an effective pathway for stable and continual reasoning improvement at test time.

arXiv:2603.03301v1 Announce Type: cross Abstract: The rapid adoption of large language models (LLMs) has created demand for faster responses and lower costs. Semantic caching, reusing semantically similar requests via their embeddings, addresses this need but breaks classic cache assumptions and raises new challenges. In this paper, we explore offline policies for semantic caching, proving that implementing an optimal offline policy is NP-hard, and propose several polynomial-time heuristics. We also present online semantic aware cache policies that combine recency, frequency, and locality. Evaluations on diverse datasets show that while frequency based policies are strong baselines, our novel variant improves semantic accuracy. Our findings reveal effective strategies for current systems and highlight substantial headroom for future innovation. All code is open source.

arXiv:2603.03304v1 Announce Type: cross Abstract: We present a concise architecture for joint training on sentences and structured data while keeping knowledge and language representations separable. The model treats knowledge graphs and hypergraphs as structured instances with role slots and encodes them into a key-value repository that a language transformer can attend over. Attention is conditioned by journey-based role transport, which unifies edge-labeled KG traversal, hyperedge traversal, and sentence structure. We outline a dual-stream architecture, hierarchical layer groups with instance-local, neighborhood, and global mixing attention, retrieval over a separate repository, and multi-task objectives spanning masked language modeling, link prediction, and role-consistency denoising. The result is an explicit, inspectable separation between linguistic context and structured knowledge, while still enabling tight alignment through cross-attention.

arXiv:2603.03305v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used to generate executable outputs, JSON objects, and API calls, where a single syntax error can make the output unusable. Constrained decoding enforces validity token-by-token via masking and renormalization, but it can distort generation when the model assigns low probability mass to valid continuations, pushing decoding toward locally valid yet semantically incorrect trajectories. We propose \emph{Draft-Conditioned Constrained Decoding (DCCD)}, a simple two-step, training-free inference procedure that decouples semantic planning from structural enforcement: an unconstrained draft is generated first, and constrained decoding is then applied, conditioned on this draft, to guarantee validity. We analyze DCCD through a KL-projection view, showing that draft conditioning increases feasible mass and reduces the cumulative "projection tax" induced by hard constraints, with an optional best-of-$K$ draft selection. Across structured reasoning benchmarks, DCCD improves strict structured accuracy by up to +24 percentage points over standard constrained decoding (e.g., 15.2\% to 39.0\% on GSM8K with a 1B model), and enables smaller model pairs to match or exceed much larger constrained baselines, yielding substantial gains in parameter efficiency.

arXiv:2603.03314v1 Announce Type: cross Abstract: Large language models (LLMs) have demonstrated remarkable and steadily improving performance across a wide range of tasks. However, LLM performance may be highly sensitive to prompt variations especially in scenarios with limited openness or strict output formatting requirements, indicating insufficient robustness. In real-world applications, user prompts provided to LLMs often contain imperfections, which may undermine the quality of the model's responses. To address this issue, previous work has primarily focused on preprocessing prompts, employing external tools or even LLMs to refine prompt formulations in advance. However, these approaches overlook the intrinsic robustness of LLMs, and their reliance on external components introduces additional computational overhead and uncertainty. In this work, we propose a Contrastive Learning-based Inverse Direct Preference Optimization (CoIPO) method that minimizes the discrepancy between the label-aligned logits produced by the model under a clean prompt and its noisy counterpart, and conduct a detailed analysis using mutual information theory. We augment the FLAN dataset by constructing paired prompts, each consisting of a clean prompt and its corresponding noisy version for training. Additionally, to evaluate the effectiveness, we develop NoisyPromptBench, a benchmark enhanced and derived from the existing PromptBench. Experimental results conducted on NoisyPromptBench demonstrate that our proposed method achieves a significant improvement in average accuracy over the current state-of-the-art approaches. The source code of CoIPO, pair-wise FLAN datasets, and NoisyPromptBench have already been released on https://github.com/vegetable-yx/CoIPO.

arXiv:2603.03315v1 Announce Type: cross Abstract: Internet memes are a powerful form of online communication, yet their nature and reliance on commonsense knowledge make toxicity detection challenging. Identifying key features for meme interpretation and understanding, is a crucial task. Previous work has been focused on some elements contributing to the meaning, such as the Textual dimension via OCR, the Visual dimension via object recognition, upper layers of meaning like the Emotional dimension, Toxicity detection via proxy variables, such as hate speech detection, and sentiment analysis. Nevertheless, there is still a lack of an overall architecture able to formally identify elements contributing to the meaning of a meme, and be used in the sense-making process. In this work, we present a semantic framework and a corresponding benchmark for automatic knowledge extraction from memes. First, we identify the necessary dimensions to understand and interpret a meme: Textual material, Visual material, Scene, Background Knowledge, Emotion, Semiotic Projection, Analogical Mapping, Overall Intent, Target Community, and Toxicity Assessment. Second, the framework guides a semi-automatic process of generating a benchmark with commonsense question-answer pairs about meme toxicity assessment and its underlying reason. The resulting benchmark M-QUEST consists of 609 question-answer pairs for 307 memes. Thirdly, we evaluate eight open-source large language models on their ability to correctly solve M-QUEST. Our results show that current models' commonsense reasoning capabilities for toxic meme interpretation vary depending on the dimension and architecture. Models with instruction tuning and reasoning capabilities significantly outperform the others, though pragmatic inference questions remain challenging. We release code, benchmark, and prompts to support future research intersecting multimodal content safety and commonsense reasoning.

arXiv:2603.03325v1 Announce Type: cross Abstract: Large language models (LLMs) have become integral to modern Human-AI collaboration workflows, where accurately understanding user intent serves as a crucial step for generating satisfactory responses. Context-aware intent understanding, which involves inferring user intentions from situational environments, is inherently challenging because it requires reasoning over both the immediate context and the user's underlying motivations that drive their behavior. Moreover, existing approaches often treat intent understanding as a static recognition task, overlooking users' accumulated intent patterns that could provide valuable references for more accurate and generalizable understanding. To address this gap, we propose IntPro, a proxy agent that learns to adapt to individual users via retrieval-conditioned intent inference. We design intent explanations that abstract how contextual signals connect to expressed intents, and store them in an individual intent history library for retrieval. We train IntPro through supervised fine-tuning on retrieval-conditioned trajectories and multi-turn Group Relative Policy Optimization (GRPO) with tool-aware reward functions, enabling the agent to learn when to leverage historical intent patterns and when to infer directly. Experiments across three diverse scenarios (Highlight-Intent, MIntRec2.0, and Weibo Post-Sync) demonstrate that IntPro achieves strong intent understanding performance with effective context-aware reasoning capabilities across different scenarios and model types.

arXiv:2603.03332v1 Announce Type: cross Abstract: Chain-of-Thought (CoT) prompting has emerged as a foundational technique for eliciting reasoning from Large Language Models (LLMs), yet the robustness of this approach to corruptions in intermediate reasoning steps remains poorly understood. This paper presents a comprehensive empirical evaluation of LLM robustness to a structured taxonomy of 5 CoT perturbation types: \textit{MathError, UnitConversion, Sycophancy, SkippedSteps,} and \textit{ExtraSteps}. We evaluate 13 models spanning three orders of magnitude in parameter count (3B to 1.5T\footnote{Assumed parameter count of closed models}), testing their ability to complete mathematical reasoning tasks despite perturbations injected at different points in the reasoning chain. Our key findings reveal heterogeneous vulnerability patterns: MathError perturbations produce the most severe degradation in small models (50-60\% accuracy loss) but show strong scaling benefits; UnitConversion remains challenging across all scales (20-30\% loss even for largest models); ExtraSteps incur minimal accuracy degradation (0-6\%) regardless of scale; Sycophancy produces modest effects (7\% loss for small models); and SkippedSteps cause intermediate damage (15\% loss). Scaling relationships follow power-law patterns, with model size serving as a protective factor against some perturbations but offering limited defense against dimensional reasoning tasks. These findings have direct implications for deploying LLMs in multi-stage reasoning pipelines and underscore the necessity of task-specific robustness assessments and mitigation strategies. The code and results are available \href{https://github.com/Mystic-Slice/CoTPerturbation}{here}.

arXiv:2603.03344v1 Announce Type: cross Abstract: Earthquake detection and seismic phase picking are fundamental yet challenging tasks in seismology due to low signal-to-noise ratios, waveform variability, and overlapping events. Recent deep-learning models achieve strong results but rely on large datasets and heavy backpropagation training, raising concerns over efficiency, interpretability, and sustainability. We propose GreenPhase, a multi-resolution, feed-forward, and mathematically interpretable model based on the Green Learning framework. GreenPhase comprises three resolution levels, each integrating unsupervised representation learning, supervised feature learning, and decision learning. Its feed-forward design eliminates backpropagation, enabling independent module optimization with stable training and clear interpretability. Predictions are refined from coarse to fine resolutions while computation is restricted to candidate regions. On the Stanford Earthquake Dataset (STEAD), GreenPhase achieves excellent performance with F1 scores of 1.0 for detection, 0.98 for P-wave picking, and 0.96 for S-wave picking. This is accomplished while reducing the computational cost (FLOPs) for inference by approximately 83% compared to state-of-the-art models. These results demonstrate that the proposed model provides an efficient, interpretable, and sustainable alternative for large-scale seismic monitoring.

arXiv:2603.03378v1 Announce Type: cross Abstract: Large language model (LLM) agents offer a promising data-driven approach to automating Site Reliability Engineering (SRE), yet their enterprise deployment is constrained by three challenges: restricted access to proprietary data, unsafe action execution under permission-governed environments, and the inability of closed systems to improve from failures. We present AOI (Autonomous Operations Intelligence), a trainable multi-agent framework formulating automated operations as a structured trajectory learning problem under security constraints. Our approach integrates three key components. First, a trainable diagnostic system applies Group Relative Policy Optimization (GRPO) to distill expert-level knowledge into locally deployed open-source models, enabling preference-based learning without exposing sensitive data. Second, a read-write separated execution architecture decomposes operational trajectories into observation, reasoning, and action phases, allowing safe learning while preventing unauthorized state mutation. Third, a Failure Trajectory Closed-Loop Evolver mines unsuccessful trajectories and converts them into corrective supervision signals, enabling continual data augmentation. Evaluated on the AIOpsLab benchmark, our contributions yield cumulative gains. (1) The AOI runtime alone achieves 66.3% best@5 success on all 86 tasks, outperforming the prior state-of-the-art (41.9%) by 24.4 points. (2) Adding Observer GRPO training, a locally deployed 14B model reaches 42.9% avg@1 on 63 held-out tasks with unseen fault types, surpassing Claude Sonnet 4.5. (3) The Evolver converts 37 failed trajectories into diagnostic guidance, improving end-to-end avg@5 by 4.8 points while reducing variance by 35%.

arXiv:2603.03387v1 Announce Type: cross Abstract: Clustering is a fundamental approach to understanding data patterns, wherein the intuitive Euclidean distance space is commonly adopted. However, this is not the case for implicit cluster distributions reflected by qualitative attribute values, e.g., the nominal values of attributes like symptoms, marital status, etc. This paper, therefore, discovered a tree-like distance structure to flexibly represent the local order relationship among intra-attribute qualitative values. That is, treating a value as the vertex of the tree allows to capture rich order relationships among the vertex value and the others. To obtain the trees in a clustering-friendly form, a joint learning mechanism is proposed to iteratively obtain more appropriate tree structures and clusters. It turns out that the latent distance space of the whole dataset can be well-represented by a forest consisting of the learned trees. Extensive experiments demonstrate that the joint learning adapts the forest to the clustering task to yield accurate results. Comparisons of 10 counterparts on 12 real benchmark datasets with significance tests verify the superiority of the proposed method.

arXiv:2603.03388v1 Announce Type: cross Abstract: Recent neural solvers have achieved strong performance on vehicle routing problems (VRPs), yet they mainly assume symmetric Euclidean distances, restricting applicability to real-world scenarios. A core challenge is encoding the relational features in asymmetric distance matrices of VRPs. Early attempts directly encoded these matrices but often failed to produce compact embeddings and generalized poorly at scale. In this paper, we propose RADAR, a scalable neural framework that augments existing neural VRP solvers with the ability to handle asymmetric inputs. RADAR addresses asymmetry from both static and dynamic perspectives. It leverages Singular Value Decomposition (SVD) on the asymmetric distance matrix to initialize compact and generalizable embeddings that inherently encode the static asymmetry in the inbound and outbound costs of each node. To further model dynamic asymmetry in embedding interactions during encoding, it replaces the standard softmax with Sinkhorn normalization that imposes joint row and column distance awareness in attention weights. Extensive experiments on synthetic and real-world benchmarks across various VRPs show that RADAR outperforms strong baselines on both in-distribution and out-of-distribution instances, demonstrating robust generalization and superior performance in solving asymmetric VRPs.