# Research Papers — 2026-06 · bundle page 249 of 260

> Full-text markdown of papers #12401–12450 of 12993 published in 2026-06, newest first. AI-readable mirrors of the full 12993-paper 2026-06 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-06/page/249
[← Previous bundle](https://aipod.dev/papers/archive/2026-06/page/248.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-06/page/250.md)

### Tiny Collaborative Inference for Occlusion-Robust Object Detection

Authors: Chieh-Tung Cheng, Mustafa Aslanov, Eiman Kanjo
Published: 2026-06-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2606.02894v2

Edge AI nodes for search and rescue are increasingly expected to run computer vision locally, yet ultra-low-end hardware imposes hard constraints on memory, compute, and inter-device communication. This work addresses occlusion-robust object detection on devices with less than 1 MB SRAM by combining an MCUNet backbone, a YOLOv2 detection head, and Lite quantisation. Two collaborative inference strategies are evaluated: feature-level fusion, concatenating intermediate feature maps, and decision-level fusion via Weighted Boxes Fusion (WBF). WBF outperforms feature-level fusion under all tested occlusion conditions, yielding gains of up to +0.2736 mAP in asymmetric scenarios. Extending fusion to three views improves accuracy further (up to +0.3827 mAP) at modest communication overhead (~1.3 KB per exchange). Hardware experiments progress from a host-assisted USB-relay baseline to a Wi-Fi peer-to-peer deployment on two Coral Dev Board Micro units, where WBF executes on-device with negligible communication energy relative to inference. In a 301.9 s autonomous session of 108 frames, fused output is produced on 61 frames versus 47 for a single board - a coverage gain of +29.8%. A decentralised federated learning feasibility note is included but not treated as a primary result, as performance remains limited under non-iid data. The results support decision-level fusion as a viable option for improving occlusion robustness in small-scale edge object detection, including host-free multi-board operation on ultra-low-end hardware.

---

### Multi-Modal Machine Learning for Breast Cancer Recurrence Prediction

Authors: Jiahao Shao, Xudong Wang, Anam Nawaz Khan, Christopher Brett, Xueping Li, Bing Yao
Published: 2026-06-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2606.02892v1

Breast cancer recurrence, a leading cause of long-term mortality among survivors, requires timely and accurate risk assessment to guide follow-up care and treatment planning. Traditional predictive models, often limited to either structured or unstructured data alone, struggle to capture the full clinical context. This study examines the impact of integrating multi-modal clinical data, including treatment records, pathology reports, and clinician notes, on recurrence prediction. By integrating a rule-based regular expression extraction mechanism with a rigorous precedence-based conflict reconciliation strategy, our approach effectively recovers definitive tumor characteristics from free-text pathology narratives to augment structured records. We also benchmark performance against commonly used feature sets from prior breast cancer studies to assess the added value of multi-modal integration. Single-source and multi-modal inputs are evaluated across a range of machine learning models. Results show that multi-modal integration consistently improves predictive accuracy compared to single-modal methods.

---

### A Nonmonotone Gradient-Based Algorithm for Symmetric Nonnegative Matrix Factorization and Graph Clustering

Authors: Ryan Swart, Johannes Brust
Published: 2026-06-01
Categories: cs.LG, math.NA, math.OC
arXiv: https://arxiv.org/abs/2606.02887v1

Symmetric nonnegative matrix factorization (Symmetric NMF) approximates a matrix as $WW^T$ with nonnegative rectangular factor $W$. It has broad applications in graph clustering and machine learning. In contrast to the NMF, projected gradient methods for the symmetric problem had been associated with slow convergence. To address this, we introduce SNMPBB, the first adaptation of nonmonotone projected Barzilai-Borwein methods to Symmetric NMF, demonstrating that gradient algorithms are significantly more effective than previously understood. We further extend SNMPBB to graph clustering using the graph Laplacian regularization (Graph-SNMPBB) and to large problems with low-rank approximations (LAI-SNMPBB). For all variants we prove global convergence to first-order stationary points and also that Barzilai-Borwein curvature information is preserved with randomized approximations. On synthetic data, SNMPBB achieves 6 times speedup over the alternative SymANLS for similar residuals, with advantages growing at higher ranks. Across six real-world clustering benchmarks, Graph-SNMPBB matches or exceeds SymANLS accuracy. Lastly, LAI-SNMPBB outperforms state-of-the-art LAI-SymPGNCG on 34 SuiteSparse matrices in both runtime and residual quality.

---

### Scalable Uncertainty Quantification for Extreme Weather Forecasting via Empirical Neural Tangent Kernels

Authors: Jose Marie Antonio Miñoza, Rex Gregor Laylo, Sebastian C. Ibañez
Published: 2026-06-01
Categories: cs.LG, cs.AI, cs.CE, math.PR, physics.ao-ph
arXiv: https://arxiv.org/abs/2606.02886v2

Deep learning weather models now match numerical weather prediction accuracy while running orders of magnitude faster, but produce deterministic forecasts without uncertainty estimates, a critical gap for high-stakes decisions during extreme weather events. This paper proposes Neural Tangent Kernel-based uncertainty quantification (NTK-UQ) using last-layer empirical features. Theoretical analysis predicts that UQ quality is architecture-dependent through two mechanisms. First, a variance collapse mechanism explains when UQ fails: when the eigenvalue truncation rank approaches the effective rank of the feature space, the GP correction term consumes nearly all prior variance, destroying discrimination between tropical cyclones and routine conditions; architectures with concentrated spectra (spectral operators) require aggressive truncation ($k \leq 10$), while attention-based models tolerate full-rank computation. Second, decomposition performance depends on the non-Gaussian, heavy-tailed structure of extreme weather: Independent Component Analysis exploits higher-order statistics (kurtosis, negentropy) to isolate heavy-tailed extreme-event features, achieving higher discrimination than singular value decomposition, which captures only second-order variance. A data-driven selection rule chooses ICA or SVD from the feature eigenspectrum concentration ratio, correctly prescribing the superior decomposition for all four evaluated architectures. Compared to split conformal prediction (the natural post-hoc baseline), NTK-UQ achieves 31--37\% sharper prediction intervals at 90\% coverage, and uniquely produces \emph{adaptive} intervals that scale with extreme event severity, which conformal prediction cannot achieve by construction. The framework requires no retraining; inference-time uncertainty requires only a single matrix-vector product per sample.

---

### Are we really tilting? The mechanics of reward guidance in flow and diffusion models

Authors: Sanjit Dandapanthula, Nicholas M. Boffi
Published: 2026-06-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2606.02884v1

Reward guidance algorithms steer a learned generative process toward the reward-tilted measure at inference time. While empirically powerful, these methods are prone to reward hacking: the guided model over-optimizes the reward at the cost of fidelity to the learned distribution. Prior work has attributed this to the complexity of neural reward functions or implicit biases in diffusion training, but its fundamental origins remain poorly understood. We show that reward hacking arises from an approximation made in most practical implementations of reward-guided diffusion -- finite-particle plug-in estimation of the Doob h-function -- even in the simplest non-trivial settings of Gaussian and Gaussian mixture targets with quadratic rewards. In closed form, we isolate two distinct failure modes of the plug-in estimator: it leads to reward hacking within each mode and it cannot select high-reward modes. We propose a closed-form reward damping schedule that corrects the within-mode bias with no additional compute, and clarify the role of best-of-n sampling in compensating for the mode selection failure. Experiments on Gaussian mixture targets, a 2D checkerboard, and FLUX.1 text-to-image generation confirm that our theoretical insights carry over to practical settings.

---

### LLM-Assisted Reranking to Operationalize Nuanced Objectives in Recommender Systems

Authors: Amir Ghasemian, Homa Hosseinmardi, Upasana Dutta, Duncan J. Watts
Published: 2026-06-01
Categories: cs.HC, cs.AI, cs.CY, cs.IR
arXiv: https://arxiv.org/abs/2606.02883v1

Recommender systems have grown from content-organization tools into sophisticated systems that shape daily behavior. By controlling what we see, they shape what we perceive, raising concerns about filter bubbles, radicalization, polarization, and social inequality. Large language models (LLMs) enable more powerful personalization, intensifying these dynamics. Yet most recommenders are tuned for engagement or limited accuracy metrics, with little attention to broader social implications, e.g. how personalization reshapes exposure in socially consequential domains. We investigate whether LLM-assisted reranking, while improving personalization, inadvertently amplifies exposure to ideologically extreme or conspiratorial political content, a risk theorized but not empirically characterized in news recommendation. Using real news-consumption histories, we rerank YouTube's sidebar candidates through zero-shot, instruction-based prompting. We compare a baseline prompt with a constrained variant that preserves topical relevance and broadens ideological exposure while reducing conspiratorial or extreme content. Without constraints, reranking strengthened personalization but increased exposure to conspiratorial and extremist material for users whose histories contained such content. Lightweight prompt-level regularization reduced promotion of extreme content and increased ideological diversity, with modest relevance loss. Synthetic experiments suggest that LLMs rerank via statistical regularities in language rather than semantic understanding of ideology, clarifying why naive prompts amplify these patterns and why regularization can reshape them. Together, our results highlight the power of LLMs to operationalize contextual nuance in high-stakes recommendation, and the need to evaluate LLM-assisted personalization beyond accuracy and treat prompt design as a value-laden rather than neutral default.

---

### Pathway-Structured Privileged Distillation for Deployable Computational Pathology

Authors: Yongxin Guo, Hao Lu, Onur Koyun, Muhammet Demir, Metin Gurcan
Published: 2026-06-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2606.02877v2

Integrating transcriptomics and histopathology can improve cancer risk modelling, yet practical use is constrained by the limited availability of RNA profiling in routine settings. Here we introduce Mixture of Pathway Experts (MoPE), a knowledge-distillation framework that reframes multimodal learning as privileged distillation for histology-only inference. MoPE is motivated by the partial observability between RNA profiles and whole-slide images: histology can capture morphology-linked consequences of certain molecular programmes, but cannot be expected to reconstruct the full transcriptomic state. MoPE encodes RNA-derived pathways and transfers the molecular supervision to pathway-indexed pathology experts through memory-usage alignment. Across diverse public benchmarks and two independent breast cancer cohorts, MoPE consistently improved WSI-only inference performance relative to baseline methods. Pathway-usage analyses and human-audited visual inspection provide bounded inspection of model behaviour and candidate morphology-linked readouts. These results support pathway-structured privileged distillation as a promising route to using molecular information during training while preserving RNA-free inference.

---

### RRISE: Robust Radius Inference via a Surrogate Estimator

Authors: Jong-Ik Park, Shreyas Chaudhari, Carlee Joe-Wong, José M. F. Moura
Published: 2026-06-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2606.02876v1

Randomized smoothing (RS) uses a smoothed classifier to provide architecture-agnostic certificates of $\ell_2$ classification robustness, but its dependence on per-input Monte Carlo (MC) sampling undermines its use in real-time systems. We argue that this cost is structural rather than fundamental, such that it can be significantly reduced by sharing information across the deployment stream. We introduce RRISE, an RS framework that compresses certification into a single forward pass through a learned surrogate. RRISE trains the surrogate against precomputed MC class-count targets via a soft-label cross-entropy loss and converts surrogate predictions into provably conservative certified radii through a one-time conformal calibration step. The resulting certificate is deployment-verifiable: whenever the calibrated radius is positive, the surrogate's prediction provably matches the smoothed classifier's and the smoothed classifier is constant on a ball of that radius around the input. Across image classification benchmarks, RRISE matches fixed-budget MC certified accuracy within $0.84$ percentage points while replacing up to $10^4$ noisy base-model evaluations per query with a single surrogate forward pass, recouping MC training cost after $\approx 10^5$ deployment queries. On CIFAR-100 and Tiny ImageNet, where the only prior offline-surrogate method collapses, RRISE achieves $1.23$ to $1.91\times$ higher certified accuracy, establishing efficient randomized smoothing as a practical path to certified robustness in repeated-deployment settings.

---

### Handoff Debt: The Rediscovery Cost When Coding Agents Take Over Interrupted Tasks

Authors: Dipesh KC, Anjila Budathoki
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02875v2

Coding-agent benchmarks evaluate whether a single uninterrupted agent can resolve a repository issue. Real software work is messier: tasks are interrupted, reassigned, reviewed, and resumed from partial states left by another agent or engineer. We study this missing dimension through handoff debt: the rediscovery cost imposed when a predecessor's work is opaque or incomplete. Our takeover protocol interrupts a coding agent at deterministic handoff points, freezes the repository, and evaluates successor agents under four handoff views: repository state only, raw trace, summary notes, and structured notes. Across 75 source tasks, the protocol generates 181 handoff-point tasks and 724 takeover runs per successor model. Across three successor models, context-bearing handoffs reduce median agent events by 20-59% and cumulative prompt tokens by 42-63% relative to repository-only takeover. Solved-rate effects are smaller and model-dependent, but efficiency gains are consistent. These findings suggest that coding-agent evaluation should report not only whether a task is solved, but also how costly that work is for another agent to resume.

---

### Handoff Debt: The Rediscovery Cost When Coding Agents Take Over Interrupted Tasks

Authors: Dipesh KC, Anjila Budathoki
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02875v1

Coding-agent benchmarks evaluate whether a single uninterrupted agent can resolve a repository issue. Real software work is messier: tasks are interrupted, reassigned, reviewed, and resumed from partial states left by another agent or engineer. We study this missing dimension through \emph{handoff debt}: the rediscovery cost imposed when a predecessor's work is opaque or incomplete. Our takeover protocol interrupts a coding agent at deterministic handoff points, freezes the repository, and evaluates successor agents under four handoff views: repository state only, raw trace, summary notes, and structured notes. Across 75 source tasks, the protocol generates 181 handoff-point tasks and 724 takeover runs per successor model. Across three successor models, context-bearing handoffs reduce median agent events by 20--59\% and cumulative prompt tokens by 42--63\% relative to repository-only takeover. Solved-rate effects are smaller and model-dependent, but efficiency gains are consistent. These findings suggest that coding-agent evaluation should report not only whether a task is solved, but also how costly that work is for another agent to resume.

---

### Neural Posterior Estimation for Stochastic Epidemic Models Using Final Outcome Data

Authors: Theodore Kypraios
Published: 2026-06-01
Categories: stat.CO, stat.ML
arXiv: https://arxiv.org/abs/2606.02874v1

Neural posterior estimation (NPE) is a simulation-based approach to Bayesian inference that trains a neural network to approximate the posterior distribution from simulated parameter - data pairs, bypassing likelihood evaluation. We apply NPE -- to our knowledge for the first time -- to stochastic susceptible-infectious-removed (SIR) epidemic models observed through final outcome data, considering both homogeneously mixing and household-structured populations. Such data arise naturally in retrospective outbreak investigations and household transmission studies, yet inference is computationally challenging: data-augmentation Markov chain Monte Carlo (MCMC) can be slow to mix in large populations and difficult to implement, while Approximate Bayesian Computation (ABC) suffers from low acceptance rates, particularly for large populations or unlikely outcomes. The discrete, low-dimensional nature of such observations makes this setting particularly well suited to NPE. We show that a logNormal posterior approximation, parameterised by a feed-forward neural network, accurately recovers reference posteriors across a range of population sizes and transmission regimes, and extends naturally to joint inference on global and local transmission rates in the household model. Once trained, the network produces approximate posterior distributions in seconds and generalises reliably to population sizes and structures not seen during training. Performance on both synthetic and real outbreak datasets is consistently strong, with results in close agreement with published analyses.

---

### Adaptive Latent Agentic Reasoning

Authors: Dongwon Jung, Peng Shi, Yi Zhang, Junshan Zhang, Muhao Chen
Published: 2026-06-01
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2606.02871v1

Large reasoning models improve performance by generating extended chain-of-thought (CoT) reasoning, but this behavior becomes inefficient when applied to LLM agents. Current LLM agents often generate verbose textual reasoning at every decision step and allocate reasoning effort nearly uniformly across turns, leading to substantial inefficiency in multi-turn agentic trajectories. We propose Adaptive Latent Agentic Reasoning (ALAR), a dual-mode framework that uses compact latent reasoning for routine turns and selectively escalates to explicit chain-of-thought when deeper deliberation is needed. ALAR learns latent reasoning by using the agent's actions as supervision anchors and is further optimized to use latent reasoning when it is sufficient for task success and reserve explicit CoT for harder decisions. Experiments on agentic search and tool-use benchmarks show that ALAR maintains comparable or better task accuracy while substantially reducing generated tokens by up to 43.6% in search and 84.6% in tool use. These results demonstrate that ALAR improves the accuracy-efficiency trade-off of LLM agents by reducing unnecessary textual reasoning while preserving explicit deliberation for harder decision steps.

---

### The Epi-LLM Framework: probing LLM behavioral priors through epidemiological agent-based models

Authors: Petra Ferencz, Ava Keeling, Tobias O'Keefe, Lorenzo Stigliano, Francesco Di Lauro, Andres Colubri, Jasmina Panovska-Griffiths
Published: 2026-06-01
Categories: cs.MA, cs.AI, q-bio.PE
arXiv: https://arxiv.org/abs/2606.02867v1

Human behaviour during epidemics affects infectious disease dynamics, but quantifying this remains deeply challenging. Here we introduce the Epi-LLM framework: a novel integration of agent-based modelling, real-life epigames, and large language models (LLMs) in which a synthetic society of agents reasons and adapts dynamically over an outbreak contact network. Comparing synthetic agent behaviour against a no-intervention SEIR baseline and human participant data from the AUIB epigame study, we find that LLM agents across four different architectures reduced peak active infections, with quarantine compliance peaking at 58-65% on day six of the 15-day simulation. A binomial generalised linear model showed that perceived health severity was the strongest predictor of quarantine behaviour ($β= 0.33, p = 0.002$), yielding a pseudo-$R^2$ of 0.055, comparable to the 0.072 observed in the human trial. LLM architecture is a key determinant of epidemic dynamics: low-variance architectures offer greater internal validity for testing behavioural rules, while high-variance models may better represent real-world decision-making. Geographic labels alone do not induce culturally differentiated behaviour; explicit attitudinal parameterisation is required. This proof-of-principle work lays the groundwork for deploying the Epi-LLM framework as a scalable, risk-free simulation environment for pandemic preparedness research.

---

### When Helping Hurts and How to Fix It: Multi-Agent Debate for Data Cleaning

Authors: Chirag Parmar, Akshat Mehta, Henglin Wu, Jagadish Ramamurthy, Shweta Medhekar
Published: 2026-06-01
Categories: cs.AI, cs.CL, cs.MA
arXiv: https://arxiv.org/abs/2606.02866v1

When does multi-agent debate help data cleaning, and when does it hurt? Across three benchmarks, four model families, and over 6,000 task-condition pairs, we find debate's effect reverses sign: it degrades generation across all four models (-1.6 to -15.5pp) through critique-induced confusion (CIC), hallucinated Critic feedback that the Generator accepts uncritically, yet improves error detection (+27.4pp F1, d=1.0). We derive a debate benefit condition: debate helps when the probability of rescuing a wrong output (Critic verification odds weighted by fixability) exceeds the probability of destroying a correct one. A factorial experiment proves adversarial separation is essential: self-verification with identical tools fails, while a separate Critic with code-execution grounding and evidence-gated generation produces the first debate configuration to significantly exceed single-agent on a generative task (+5.3pp, p<0.05). The condition correctly predicts all nine task types and generalizes with zero false positives across 19 published comparisons in seven domains.

---

### Don't Gamble, GAMBLe: An Analytical Framework for AI-Driven Research Systems

Authors: Marquita Ellis, Paul Castro
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02863v2

AI-Driven Research Systems (ADRS) -- systems coupling LLMs with automated evaluation to discover algorithms, proofs, and designs -- are being optimized and adopted across domains, but the tools to analyze them have not kept pace. ADRS performance depends on component interactions that are poorly understood, expensive to explore, and (as we show) not well captured by standard convergence guarantees. These guarantees rely on structural assumptions that do not hold under the ADRS process we formalize. We introduce GAMBLe, a framework that decomposes ADRS behavior into four parameters (generator $G$, assessor $\mathcal{A}$, discovery mechanism $\mathcal{M}$, budget $B$) and one compositional object, the effective landscape $L_{\text{eff}} = \mathcal{A} \circ G$, which reveals that distinct generator-assessor pairs induce structurally different per-problem optimization landscapes. We exercise the framework on 760+ replicated runs (>46,000 iterations) spanning generators from single LLMs to dynamically-adaptive ensembles, mechanisms from greedy selection to co-evolutionary meta-search, and three NP-hard problems whose assessors range from continuous scoring to cliff functions. The experiments reveal no total ordering of generators or mechanisms: frontier models can underperform open-source alternatives and the simplest mechanism sometimes outperforms state-of-the-art meta-search. Results show that even under limited budgets (60 iterations per run), the right component choices can improve performance by 13-67% and search efficiency by 6-39x.

---

### Toward a Modular Architecture for Embedded AI Agent Systems at the Edge

Authors: Marcus Rüb, Michael Gerhards
Published: 2026-06-01
Categories: cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2606.02862v1

The rise of Large Language Models (LLMs) has enabled agentic AI capable of complex reasoning and tool use; however, deploying such autonomy in pervasive computing environments remains challenging due to the strict memory and energy constraints of embedded microcontrollers. Existing frameworks typically assume server-class resources or continuous connectivity, leaving a gap for deeply embedded systems. This paper proposes a modular reference architecture for Embedded Agent Systems that bridges the divide between deterministic real-time control and agentic intelligence. We introduce a tiered design that decouples On-Device Agents - executing highly compressed neural networks and rule-based logic for low-latency, privacy-critical tasks - from Cloud-Augmented Agents that leverage Small Language Models (SLMs) for higher-level reasoning and planning. A key contribution is the integration of a cross-cutting Governance Layer, ensuring observability, policy enforcement, and safety across distributed fleets of autonomous devices. Rather than presenting purely empirical benchmarks, we analyze architectural design principles and trade-offs regarding latency, energy, and reliable execution in resource-constrained environments.

---

### Forgetting is Not Erasure: Recovering Latent Knowledge via Transport Keys

Authors: Archie Chaudhury
Published: 2026-06-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2606.02860v1

Catastrophic forgetting is often framed as a representational problem: after sequential training, a model appears to lose the features that supported performance on earlier tasks. We challenge the stronger form of this view. Across controlled continual-learning settings, we find that a significant portion of apparent forgetting can be attributed to interface drift between internal stages rather than permanent erasure of task-relevant computation. We study this phenomenon through a stitched evaluation protocol that combines early computation from a post-update network with late computation from its predecessor, optionally mediated by a compact, task-specific transport key. We describe transport keys at a systems level as compact interface-alignment operators estimated from a small set of paired anchor activations and evaluated through model stitching. On split CIFAR-100 with a ResNet-style network, transport keys recover most of the original Task A performance after sequential training on Task B. On a compact vision transformer, we observe a similar recovery pattern. These results suggest that continual learning may require better mechanisms for indexing and re-accessing latent computations, not only methods that prevent weight change.

---

### Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions

Authors: Zhenting Qi, Huangyuan Su, Ao Qu, Chenyu Wang, Yu Yao, Han Zheng, Kushal Chattopadhyay, Guowei Xu, Zihan Wang, Weirui Ye, Vijay Janapa Reddi, Ju Li et al.
Published: 2026-06-01
Categories: cs.CL, cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2606.02859v1

How can a population of agents self-orchestrate and self-adapt into stronger collective intelligence without centralized control? Inspired by Friedrich Hayek's economic theory of decentralized coordination in markets, we study this question through an agent economy in which agents compete via auctions for the right to act, exchange payments, and accumulate wealth from environmental rewards. These simple economic signals induce decentralized credit assignment, driving planning without global orchestration or explicit communication protocols. The population evolves through economic selection: effective agents accumulate wealth and are mutated via exploitation, while ineffective ones go bankrupt and are replaced via exploration. We show that, initialized with weak agents, the economy produces emergent multi-step reasoning strategies and outperforms stronger monolithic baselines across five agentic tasks, including mathematical reasoning, financial research, scientific research, accelerator design, and distributed-system optimization. We further provide theoretical insights into how economic dynamics shape agent behaviors, linking local incentives to long-term global performance. Our results suggest a new path to multi-agent intelligence: rather than engineering coordination, we can design decentralized incentive structures under which it automatically emerges.

---

### GRZO: Group-Relative Zeroth-Order Optimization for Large Language Model Fine-Tuning

Authors: Liyan Tan, Yequan Zhao, Yifan Yang, Ruijie Zhang, Xinling Yu, Zheng Zhang
Published: 2026-06-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2606.02857v2

Zeroth-order (ZO) optimization is a memory-efficient alternative to backpropagation for fine-tuning large language models, but its deployment is limited by the high variance of gradient estimation. We propose GRZO, a Group-Relative Zeroth-Order optimizer that draws one pseudo-independent perturbation per mini-batch example and aggregates the per-example losses through group-relative normalization, raising the effective gradient-direction count from one to the batch size at no additional forward cost while preserving inference-level memory. We prove that GRZO is directionally unbiased with variance shrinking proportionally to the batch size, yielding a tighter nonconvex convergence bound than MeZO. Across RoBERTa-large, Llama3-8B, and OPT-13B over multiple tasks, GRZO improves average accuracy on Llama3-8B by $+3.0$ over MeZO at $23\%$ lower peak GPU memory; as a drop-in replacement for the MeZO core, it lifts sparse, low-rank, and quantized ZO variants by $+6.0$ on average.

---

### GRZO: Group-Relative Zeroth-Order Optimization for Large Language Model Fine-Tuning

Authors: Liyan Tan, Yequan Zhao, Yifan Yang, Ruijie Zhang, Xinling Yu, Zheng Zhang
Published: 2026-06-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2606.02857v1

Zeroth-order (ZO) optimization is a memory-efficient alternative to backpropagation for fine-tuning large language models, but its deployment is limited by the high variance of gradient estimation. We propose GRZO, a Group-Relative Zeroth-Order optimizer that draws one pseudo-independent perturbation per mini-batch example and aggregates the per-example losses through group-relative normalization, raising the effective gradient-direction count from one to the batch size at no additional forward cost while preserving inference-level memory. We prove that GRZO is directionally unbiased with variance shrinking proportionally to the batch size, yielding a tighter nonconvex convergence bound than MeZO. Across RoBERTa-large, Llama3-8B, and OPT-13B over multiple tasks, GRZO improves average accuracy on Llama3-8B by $+3.0$ over MeZO at $23\%$ lower peak GPU memory; as a drop-in replacement for the MeZO core, it lifts sparse, low-rank, and quantized ZO variants by $+6.0$ on average.

---

### RESCAST-100K: A Comprehensive Dataset for Cross-Domain Residential Load and Indoor Temperature Forecasting

Authors: Jainam Dhruva, Yousaf Raza, A. B. Siddique, Simone Silvestri
Published: 2026-06-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2606.02852v1

Accurate short-term forecasting of residential energy load and indoor temperature is essential for home energy management systems, grid-level demand response, and community energy efficiency efforts. Domain adaptation and transfer learning have shown promise for improving forecasting accuracy under data heterogeneity and scarcity commonly seen in residential settings. However, progress is limited by the lack of comprehensive residential datasets: existing benchmarks are narrow in target coverage and rarely support structured cross-domain evaluation. We introduce RESCAST-100K, a large-scale residential forecasting benchmark for studying cross-domain generalization. It provides a configuration-driven interface that instantiates source and target domains along interpretable axes, including geography, climate zone, wall construction, and heating equipment, enabling systematic evaluation of transfer learning, domain adaptation, and zero-shot generalization under controlled domain shifts. The benchmark covers approximately 100,000 EnergyPlus-simulated U.S. homes derived from ResStock, with 15-minute time series for three coupled targets per home: total load, HVAC load, and indoor temperature. These are paired with weather channels, HVAC setpoints, and over 40 static building covariates. RESCAST-100K also integrates five real-world residential datasets under a unified schema, supporting sim-to-real evaluation on the same tasks. We benchmark recurrent, attention-based, and MLP-mixer architectures for zero-shot performance across domains, missing-input conditions, and forecasting tasks. Cross-attention and MLP-mixer models consistently outperform recurrent and classical transformer baselines under domain shift. RESCAST-100K is intended to aid the machine learning and building analytics communities advance cross-domain residential forecasting at home, community, and grid scale.

---

### A Systematic Evaluation of Current Architectures in Wind Power Forecasting

Authors: Vinicius Bortolini, Gilson Adamczuk Oliveira, Erick Oliveira Rodrigues, Matheus Henrique Dal Molin Ribeiro
Published: 2026-06-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2606.02849v1

Interval wind speed forecasting is essential for the efficient integration of wind energy into power systems, as it accounts for the inherent uncertainty of wind resources. This study presents a systematic literature review focused on hybrid approaches to interval forecasting of wind generation, exploring the combination of deep learning, modal decomposition, and statistical methods. To guide the paper selection, Latent Dirichlet Allocation (LDA) was applied for topic modeling, enabling the identification of patterns and research trends. The findings emphasize that integrating hybrid models with decomposition techniques-such as Variational Mode Decomposition (VMD) and Ensemble Empirical Mode Decomposition (EEMD)-enhances forecast accuracy and reliability by narrowing prediction intervals without compromising coverage. Regarding interval construction, most studies adopt a dual-model strategy, independently forecasting the lower and upper bounds. Input data are commonly decomposed using techniques like EMD, EEMD, or VMD, which extract frequency-based components. These components serve as inputs to models such as LSTM or ELM, trained separately for each bound. This approach allows for targeted modeling of uncertainty, improving flexibility and precision, Interval quality is typically evaluated through metrics that balance coverage and interval width. The review also highlights challenges, including the lack of standardized evaluation metrics, computational complexity, and limited real-world validation. Overall, the study reinforces the value of interval forecasting for wind energy operations and offers insights for advancing model robustness and decision-making.

---

### Spectral-Progressive Thought Flow for Lightweight Multimodal Reasoning

Authors: Yixian Shen, Zhiheng Yang, Qi Bi, Changshuo Wang, Shuai Wang, Jia-Hong Huang, George Floros, Prayag Tiwari, Anuj Pathania
Published: 2026-06-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2606.02842v1

Multimodal spatial reasoning often relies on long chains of intermediate textual and visual thoughts, where accumulating visual tokens and dense cross-modal attention incur substantial computation and memory overhead. To address this challenge, we propose Spectral-Progressive Thought Flow (SpecFlow), a novel lightweight multimodal spatial reasoning framework that represents intermediate visual thoughts in a fixed-size discrete cosine space. By exploiting strong energy compaction, SpecFlow preserves global layout and relational structure while introducing high-frequency details only when increased spatial precision is required. To align visual state evolution with linguistic intent, classifier-free guidance enables autoregressive textual thoughts to steer flow-based updates of the visual workspace/state without expanding the context. As a result, SpecFlow maintains a bounded visual workspace whose updates depend only on the current visual state and accumulated textual trace, enabling long-horizon inference with stable latency and memory usage independent of reasoning depth. Empirical results show that SpecFlow achieves competitive or superior reasoning performance while reducing computation and KV cache costs by up to 2.1 times.

---

### Learning Coherent Representations: A Topological Approach to Interpretability

Authors: Sigurd Gaukstad, Melvin Vaupel, Valdemar Kargård Olsen, Erik Hermansen, Benjamin Dunn
Published: 2026-06-01
Categories: cs.LG, math.AT
arXiv: https://arxiv.org/abs/2606.02841v1

Deep neural networks learn representations where individual features often lack interpretable meaning; a single neuron may activate for scattered, unrelated inputs. We introduce coherence, a geometric property inspired by neural coding in the brain, where neurons like grid cells and head direction cells respond to contiguous regions of state space. A non-negative matrix is coherent if each row (sample) attends to geometrically clustered columns (features) and vice versa, and in addition every sample is well described by some feature and every feature is needed by some sample. We prove that coherent matrices induce a bounded interleaving between the Vietoris-Rips filtrations of samples and features, guaranteeing that both spaces share compatible topological structure. This geometric constraint facilitates interpretability. For example, if data lies on a circle, coherent features must tile that circle into contiguous arcs. We introduce Coh, a differentiable objective function based on Fréchet variance that enforces coherence during training. Unlike sparsity, which bounds how many samples a feature activates on, coherence bounds which samples, requiring geometric connectivity rather than only rarity. This yields not just interpretable features but an interpretable feature space. We validate Coh in an auto-encoder using synthetic and rotated MNIST datasets and in a token embedding of BERT using language data.

---

### Self-Regulation through Communication in Evolved Neural Agents

Authors: Joshua Nunley
Published: 2026-06-01
Categories: q-bio.PE, cs.MA, cs.NE, nlin.AO
arXiv: https://arxiv.org/abs/2606.02840v1

Communication is typically understood as indication: signals that transfer information from sender to receiver. We present a minimal predator avoidance task in which pairs of evolved CTRNN agents use communication for robust survival, and in which agents hear their own vocalizations, as in natural systems. Across 112 perfect-fitness agents from over 2,000 evolutionary runs, three dominant strategies emerge (accounting for 81% of agents): safety calling (39%), where agents signal from safe cover; alarm indication (22%), where agents vocalize when a threat is present without relying on self-hearing; and self-regulatory calling (20%), where agents depend on hearing their own call to sustain escape behavior. Self-hearing dependency is common among agents that call during an active threat (47%), but rare among agents that call only after reaching safe cover (10%; p < 10^-4). This pattern is consistent with a difference in causal order: safety callers act then communicate, while self-regulatory callers communicate in order to act. Removing self-hearing selectively impairs self-regulatory callers (fitness 0.40) while safety callers remain functional (0.90; p < 10^-9). These results show that communication can evolve to serve the caller's own behavioral regulation, not just information transfer to others.

---

### Fixing FOLIO and MALLS: Verified Annotations and an LLM-assisted Framework to Focus Human Relabeling

Authors: Andrea Brunello, Cristian Curaba, Luca Geatti, Michele Mignani, Angelo Montanari, Nicola Saccomanno
Published: 2026-06-01
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2606.02837v2

Accurate translation from Natural Language to First-Order Logic (NL-to-FOL) underpins neurosymbolic AI systems and Natural Language Inference (NLI), making the quality of NL-to-FOL benchmarks essential---yet these datasets have never been rigorously audited. Our first contribution is to present a systematic human inspection of the validation split of \textsf{FOLIO} and a subset of \textsf{MALLS} test instances, finding that approximately 42.5\% and 42\% of entries, respectively, contain incorrect FOL formalizations (i.e., ground truth labels), with additional rates of ambiguous NL sentences (17.8\% and 51\%) and incorrect NLI labels in \textsf{FOLIO} (8.4\%). Our second contribution is to develop and release corrected ground truths for such datasets, showing that annotation errors distort model evaluation on a reference benchmark task: testing three state-of-the-art LLMs (Gemma~4 31B-it, Qwen3-30B-A3B, and GPT-4o-mini) with the corrected ground truths yields accuracy gains from +11 to +23 percentage points. Motivated by these findings, we propose an LLM-based framework to support humans in manual reviewing NL-to-FOL datasets. By directing reviewers toward the most error-prone instances, we empirically show that it is possible to achieve 90\% dataset accuracy after reviewing fewer than 20\% of instances, compared to over 76\% required by unguided review. We release all human-verified annotations and the code for our framework.

---

### Fixing FOLIO and MALLS: Verified Annotations and an LLM-assisted Framework to Focus Human Relabeling

Authors: Andrea Brunello, Cristian Curaba, Luca Geatti, Michele Mignani, Angelo Montanari, Nicola Saccomanno
Published: 2026-06-01
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2606.02837v1

Accurate translation from Natural Language to First-Order Logic (NL-to-FOL) underpins neurosymbolic AI systems and Natural Language Inference (NLI), making the quality of NL-to-FOL benchmarks essential -- yet these datasets have never been rigorously audited. Our first contribution is to present a systematic human inspection of the validation split of \textsf{FOLIO} and a subset of \textsf{MALLS} test instances, finding that approximately 39% and 36% of entries, respectively, contain incorrect FOL formalizations (i.e., ground truth labels), with additional rates of ambiguous NL sentences (16.4% and 48%) and incorrect NLI labels in \textsf{FOLIO} (8.4%). Our second contribution is to develop and release corrected ground truths for such datasets, showing that annotation errors distort model evaluation on a reference benchmark task: testing three state-of-the-art LLMs (Gemma~4 31B-it, Qwen3-30B-A3B, and GPT-4o-mini) with the corrected ground truths yields accuracy gains from +9 to +22 percentage points. Motivated by these findings, we propose an LLM-based framework to support humans in manual reviewing NL-to-FOL datasets. By directing reviewers toward the most error-prone instances, we empirically show that it is possible to achieve 90% dataset accuracy after reviewing fewer than 24% of instances, compared to over 70% required by unguided review. We release all human-verified annotations and the code for our framework.

---

### Thinking Past the Answer: Evaluating Harmful Overthinking in Large Reasoning Models

Authors: Simone Caldarella, Davide Talon, Rahaf Aljundi, Elisa Ricci, Massimiliano Mancini
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02835v1

Large Reasoning Models (LRMs) improve performance by generating explicit intermediate reasoning traces through increased test-time compute, yet the assumption that longer reasoning is consistently beneficial remains under-examined. While recent evidence shows that additional reasoning can lead models to overthink, we ask: "Once a model has reached the correct answer, does further reasoning refine the solution, or deviate from it?" To study the dynamics after correctness, we introduce a prefix-level trajectory evaluation protocol grounded in reasoning sufficiency, defining the minimum reasoning budget required for a model to first generate the correct answer. This allows us to disentangle verbose overthinking, where additional reasoning is redundant but harmless, from harmful overthinking, where continued reasoning destabilizes an already-correct trajectory. Starting from multimodal benchmarks, we find that many instances considered reasoning-intensive require surprisingly little reasoning. Moreover, stopping at the first correct prefix improves accuracy over standard reasoning up to 21%, revealing that current models are limited not only by their ability to reason, but also by their inability to stop at the right time. Furthermore, while common efficiency strategies like early stopping substantially reduce verbose overthinking (up to 50%), they fail to mitigate harmful overthinking. Failure analysis reveals that correctness deviations are mainly driven by logical drift and visual reinterpretation. Finally, we show that our findings generalize to language-only reasoning benchmarks, highlighting harmful overthinking as a broader reliability risk. Code available at https://simonecaldarella.github.io/thinking-past-the-answer.

---

### Large Byte Model: Teaching Language Models About Compiled Code

Authors: Florian Störtz, Catalin-Andrei Stan, Alexandru Dinu, Sandra Servia-Rodríguez, Mihaela Gaman, Calin Miron, Edward Raff
Published: 2026-06-01
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2606.02834v1

Malware analysis starts with the raw bytes of an executable program, and tools to "lift" these to higher-level representations, such as assembly, are expensive and subject to error. Large Language Models (LLMs) cannot process raw byte representations and answer questions about them. To this end, we present the first byte-native LLM. Based on a vocabulary expansion technique using a bespoke byte tokenizer, such a model is capable of responding to complex questions about malware binaries, with accuracies ranging from 69% for malware family classification to 98% for architecture classification. Our findings indicate that providing domain knowledge during training is essential for this application -- off-the-shelf models lack both accuracy and insight. We've deployed this emerging solution to a limited number of analysts to gather feedback for further improvements.

---

### An Exploration of Collision-based Enemy Morphology Generation

Authors: Johor Jara Gonzalez, Matthew Guzdial
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02832v1

Despite a great deal of prior research into Procedural Content Generation (PCG), relatively little prior work has explored generating enemies for video games. In particular, there is almost no work on generating enemy morphologies, the basic body plan or collision information for in-game enemies, despite the existence of related morphology generation work in robotics. In this paper, we explore three different novel approaches to generate enemy morphologies based on player collision information. We found that each approach provides different strengths and weaknesses, but all had equivalent or better performance than an evolutionary baseline adapted from prior robotics morphology work.

---

### Principled Reflection Separation via Nonlinear Superposition and Feature Interaction

Authors: Qiming Hu, Mingjia Li, Yuntong Li, Xiaojie Guo
Published: 2026-06-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2606.02831v1

Single-image reflection separation is fundamentally challenged by the entanglement of transmission and reflection layers under complex image formation processes. Existing approaches largely rely on simplified assumptions or independent modeling, limiting their ability to handle real-world scenarios. In this work, we revisit the problem from a unified perspective and identify a key issue of existing approaches, i.e., the widely adopted linear composition model in the sRGB domain fails to capture the nonlinear coupling introduced by real-world image signal processing pipelines. To address this, we introduce a learnable nonlinear superposition model that more faithfully characterizes layer interactions and improves decomposition fidelity. Building upon this formulation, we propose a generalized dual-stream interactive framework that explicitly models bidirectional dependencies between transmission and reflection through feature exchange. This framework unifies activation-, gating-, and attention-based interaction mechanisms, and is compatible with both CNN and Transformer backbones. Extensive experiments on diverse real-world benchmarks demonstrate that the proposed approach achieves superior performance with strong generalization capability. More importantly, our study reveals that reflection separation is not about undoing a linear mixture, but about learning nonlinear formation and interaction}, offering new insights into the design of principled image decomposition models. Code and models are publicly available at https://mingcv.github.io/DIRS-Page.

---

### Mitigating Spurious Correlations with Memorization-Guided Dataset De-Biasing

Authors: Arda Fazla, Abolfazl Hashemi
Published: 2026-06-01
Categories: cs.LG, math.OC
arXiv: https://arxiv.org/abs/2606.02830v1

Real-world datasets often contain spurious correlations that are not causally related to the target label. When such correlations dominate the majority of training samples, models tend to rely on them, leading to misclassification of minority samples that do not exhibit the same spurious patterns. While a potential approach is to select subsets of data to better represent the minority samples, this may require access to group labels, which are typically unknown. Furthermore, as we demonstrate, widely used sample scoring functions in the invariant subset or coreset selection literature largely depend on spurious features and therefore fail to accurately capture the importance or difficulty of core, causally relevant features. Accordingly, we propose to mitigate spurious correlations by developing a two-stage sample scoring function that disentangles the learning dynamics of core and spurious features and evaluates their difficulty separately. Based on our proposed metric, we introduce a new algorithm to find and prioritize informative samples both with and without spurious correlations. Extensive experiments demonstrate that a standard ERM model trained on our selected samples achieves superior performance compared to state-of-the-art debiasing techniques, while requiring as little as 10\% of the original training data.

---

### Qift: Shift-Friendly No-Zero W2 Post-Training Quantization for Rotated W2A4/KV4 LLM Inference

Authors: Chi-Wei Huang, Chia-Chi Tsai
Published: 2026-06-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2606.02823v1

Two-bit weight quantization is attractive for memory-efficient LLM inference, but the standard W2 level set {-2,-1,0,+1} often collapses under aggressive W2A4/KV4 settings. We study the scalar level-set geometry of two-bit weights in a Hadamard-rotated quantization pipeline. Conventional asymmetric W2 substantially improves over the standard level set, indicating that W2A4 failure is not only a bit-width problem but also a reconstruction-level problem. Across all 224 linear modules in each of LLaMA-2-7B and LLaMA-3.1-8B, pretrained weights are already nearly zero-centered, while Hadamard rotation primarily Gaussianizes their standardized shape: excess kurtosis and Q-Q error drop by orders of magnitude. Based on this approximate zero-centered Gaussian-like source model, we propose Qift, a fixed no-zero W2 level set for rotated W2A4/KV4 inference. The main level set is {+/-0.5, +/-1.5}, equivalently {+/-1, +/-3} under a half-scale reparameterization; a power-of-two variant uses {+/-1, +/-4} for sign-and-shift decoded weight application. Qift redesigns the fixed two-bit code-to-level mapping and is training-free, learned-codebook-free, group-grid-free, and zero-point-free, retaining the standard per-channel scale. A scale-invariant ratio analysis identifies an effective inner/outer centroid ratio range of 0.25 to 0.33, explaining why mirror no-zero (MNZ), Lloyd, NF2, and PoT-MNZ perform well while {+/-1, +/-2} does not. On both models, the no-zero level sets consistently improve pure W2A4 perplexity, L-layer mixed W2/W4 perplexity, downstream accuracy, and GPTQ residual behavior over the standard W2 level set. At L=16 mixed precision, they substantially narrow the gap to W3A4 while keeping half of the transformer layers at two-bit precision, giving a simple, source-aware, and deployment-friendly alternative to more complex learned W2 codebooks.

---

### Which Defense Closes Which Threat? Attributing OWASP-LLM-Top-10 Coverage and Its Brittleness Under Paraphrasing

Authors: Alexandre Cristovão Maiorano
Published: 2026-06-01
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2606.02822v1

Production LLM applications stack several defense families -- refusal-phrase filters, token-budget controls, model allowlists, rate limits, tool-registry authentication -- yet existing breach-and-attack-simulation (BAS) benchmarks report a single aggregate coverage number, hiding which family closes which threat. We measure attribution. We add four OWASP-LLM-Top-10-aware agents to a 21-agent baseline scanner and target a lattice of four synthetic LLM endpoints: $L_0$ (no defenses), $L_1$ (refusal-only), $L_2$ (budget-only), and $L_3$ (full stack). $L_1$ and $L_2$ are sibling single-axis ablations, not subsets of each other; $L_3$ is their union plus tool-registry authentication and credential scrubbing. Across $N=10$ replications, the per-OWASP finding count is clean: refusal alone removes all LLM01 (jailbreak) and LLM07 (system-prompt leakage) findings; budget alone removes all LLM02 (sensitive-info disclosure) and LLM10 (unbounded consumption) findings by terminating multi-step sequences; LLM06 (excessive agency) requires the full stack. We probe brittleness under paraphrasing: with 300 Gemini-generated paraphrases ($K=5$ over a 60-template brittleness corpus), $L_1$ refusal block rate falls 15 pp on LLM01 and 25 pp on LLM07. A fifth target, $L_4$-real, swaps the stub backend for Gemini-2.5-flash behind the same $L_3$ regex and matches $L_1$ exactly, indicating no measurable alignment contribution beyond the regex (not a general claim about alignment). Budget controls show no drop (0 pp once the rate-limit floor is factored out). A refusal whitelist that clears a static benchmark can be defeated by an LLM-driven paraphraser without changing attack intent; a budget control resists the same mutation.

---

### Do Neural Retrievers Prefer Certain Documents? Evidence of Learned Relevance Priors

Authors: Francisco Valentini, Edgar Altszyler, Martin Fajcik
Published: 2026-06-01
Categories: cs.IR, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2606.02814v1

Neural retrievers are trained to estimate query-document relevance from annotated query-document pairs. Yet annotation protocols may not purely reflect relevance: they select only a subset of documents for labeling, and this selection can favor certain document types over others. We investigate whether supervised bi-encoder retrievers implicitly learn a document-level relevance prior: a query-independent signal encoded in their representation space as a side effect of training on annotated data. We estimate this prior by training simple classifiers on frozen document embeddings and evaluate three state-of-the-art retrievers across multiple IR benchmarks. We find that supervised neural retrievers encode relevance priors that generalize to unseen documents and are consistent across models. These priors create a findability gap: documents with lower prior are systematically harder to retrieve, even when genuinely relevant. This effect appears in supervised dense retrievers but is weaker and less consistent in BM25, and it persists under controlled matched-document comparisons. Using LLM-based explanations, we find that judged-relevant documents tend to be comprehensive, self-contained summaries of mainstream topics, while niche, fragmentary, or highly technical content is often left unjudged. Retrievers internalize this bias, ranking documents with these favored features higher than documents that lack them, independently of their actual relevance. Our findings expose a structural limitation of supervised retrieval: models trained on annotated data do not just learn relevance, but also the implicit document preferences in their training data.

---

### Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection

Authors: Sihang Zeng, Matthew Thompson, Ruth Etzioni, Meliha Yetisgen
Published: 2026-06-01
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2606.02812v1

Modeling patient trajectories from longitudinal electronic health records (EHRs) requires reasoning over sparse, noisy, and long-context multimodal sequences. Existing LLM-based multi-agent systems address context length but process patients in isolation, failing to mirror how clinicians leverage accumulated experience from similar prior cases. We present Traj-Evolve, a self-evolving multi-agent system with two complementary evolving mechanisms. First, an Experience Pool (ExPool) acts as a non-parametric memory, indexing rejection-sampled reasoning traces to retrieve similar patients as few-shot contexts. Second, multi-agent reinforcement learning (MARL) via reward-ranked fine-tuning parametrically optimizes inter-agent and agent-memory collaboration. A leave-one-out cross-retrieval strategy unifies the two, aligning training- and inference-time behavior under retrieval augmentation. On a lung cancer prediction task utilizing up to five years of multimodal EHRs, Traj-Evolve outperforms 9 strong baselines on the overall population and a challenging never-smoker population. Analysis of the evolving dynamics highlights three key findings: (1) expanding the ExPool shifts optimal retrieval from diverse to specific samples; (2) under MARL, the manager agent's prediction loss converges quickly while the worker agents' temporal reasoning continues to benefit from more verified patients; and (3) the two mechanisms are complementary on the predicted risk, where ExPool improves specificity while MARL improves sensitivity.

---

### Automated Report-Derived Oncology VQA Benchmark for Evaluating Vision-Language Models on 3D Medical Imaging

Authors: Bo Liu, Hanxue Gu, Xiangru Li, Zheren Zhu, Jacob Ellison, Kang Wang, Janine M. Lupo, Yang Yang, Hui Lin
Published: 2026-06-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2606.02809v1

Evaluating vision-language models (VLMs) on medical images requires benchmarks that are clinically grounded, scalable, and controlled for evaluation confounds. Existing public benchmarks are limited in scale, manually annotated, or potentially leaked into VLM pretraining corpora. We present an automated agent-driven pipeline that generates multiple-choice VQA datasets directly from paired private radiology reports and 3D oncology imaging, producing two complementary question types: RADS-style questions deterministically derived from clinician-defined reporting schemas, and radiology report-derived questions generated by an LLM from radiologist findings and verified against the source report. Applied to four in-house cancer cohorts, the pipeline yields an instance-contamination-controlled benchmark without per-question human annotation. Zero-shot evaluation of six VLMs reveals no dominant model and substantial headroom across all cells. A blind ablation reveals that visual reliance is highly dataset-specific: liver Report-derived questions genuinely require the image, while Lung CT is essentially solvable without it - the leading closed model exceeds its sighted accuracy on Lung CT when blinded - indicating that even private clinical data does not guarantee a contamination-controlled read of visual capability. The pipeline is released as an open agent skill for in-house redeployment.

---

### Translating Classical Poetry into Modern Prose

Authors: Chalamalasetti Kranti, Sowmya Vajjala
Published: 2026-06-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2606.02806v3

We built a dataset for the task of poem-to-prose translation from 13th-17th Century Telugu classical poetry to contemporary Telugu and English prose, which we call Padyam2Gadyam. The dataset consists of 600 poems and their human-verified Telugu and English prose translations. We evaluated 2 machine translation systems and 5 contemporary Large Language Models (LLMs) on their ability to do zero-shot poem-to-prose translation into Telugu and English using this dataset. Our results indicate that while the general purpose LLMs are better than the machine translation systems, there are systematic issues with the generation and evaluation of prose translation in both languages across systems.

---

### Translating Classical Poetry into Modern Prose

Authors: Chalamalasetti Kranti, Sowmya Vajjala
Published: 2026-06-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2606.02806v2

We introduce Padyam2Gadyam a dataset for the task of poem-to-prose translation from 13th-17th Century Telugu Classical Poetry to contemporary Telugu and English prose. The dataset consists of 600 poems and their human-verified Telugu and English prose translations. We evaluated 2 machine translation systems and 5 contemporary Large Language Models (LLMs) on their ability to do poem-to-prose translation into Telugu and English using this dataset. Our results indicate that while the general purpose LLMs are better than the machine translation systems for this task, there are systematic issues with the generation and evaluation of prose translation in both languages across systems.

---

### ChatHealthAI: Aligning Electronic Health Record Representations with Large Language Models for Grounded Clinical Reasoning

Authors: Bo-Hong Wang, Baicheng Peng, Ruilin Wang, Jun Bai, Ziyang Song, Yue Li
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02802v2

Large language models (LLMs) exhibit strong natural-language reasoning abilities for clinical decision support, but struggle to effectively model structured longitudinal electronic health records (EHRs). In contrast, EHR foundation models can learn predictive patient representations, yet lack interpretable language-based reasoning. To bridge this gap, we propose ChatHealthAI, a multimodal reasoning framework that aligns structured EHR representations from a pretrained EHR foundation model with the semantic space of a frozen LLM through a task-aware resampler. By integrating longitudinal patient representations with refined clinical event descriptions, ChatHealthAI enables clinically grounded natural-language reasoning while maintaining accurate patient prediction. We evaluated ChatHealthAI on three clinical predictive tasks from the EHRSHOT benchmark. Results show that ChatHealthAI improves reasoning quality and interpretability while preserving competitive predictive performance. These findings highlight the potential of integrating EHR foundation models with pretrained LLMs for interpretable clinical prediction.

---

### Cosmos 3: Omnimodal World Models for Physical AI

Authors: NVIDIA, :, Aditi, Niket Agarwal, Arslan Ali, Jon Allen, Martin Antolini, Adeline Aubame, Alisson Azzolini, Junjie Bai, Maciej Bala, Yogesh Balaji et al.
Published: 2026-06-01
Categories: cs.CV, cs.AI, cs.LG, cs.MM, cs.RO
arXiv: https://arxiv.org/abs/2606.02800v4

We introduce Cosmos 3, a family of omnimodal world models designed to jointly process and generate language, image, video, audio, and action sequences within a unified mixture-of-transformers architecture. By supporting highly flexible input-output configurations, Cosmos 3 seamlessly unifies critical modalities for Physical AI -- effectively subsuming vision-language models, video generators, world simulators, and world-action models into a single framework. Our evaluation demonstrates that Cosmos 3 establishes a new state-of-the-art across a diverse suite of understanding and generation tasks, demonstrating omnimodal world models as scalable, general-purpose backbones for embodied agents. Our post-trained Cosmos 3 models were ranked as the best open-source Text-to-Image and Image-to-Video models by Artificial Analysis, and the best policy model by RoboArena at the time the technical report was written. To accelerate open research and deployment in Physical AI, we make our code, model checkpoints, curated synthetic datasets, and evaluation benchmark available under the Linux Foundation's OpenMDW-1.1 License at https://github.com/nvidia/cosmos and https://huggingface.co/collections/nvidia/cosmos3. The project website is available at https://research.nvidia.com/labs/cosmos-lab/cosmos3.

---

### BehaviorBench: Modeling Real-World User Decisions from Behavioral Traces

Authors: Liangwei Yang, Jielin Qiu, Zixiang Chen, Ming Zhu, Juntao Tan, Zhiwei Liu, Wenting Zhao, Zhujun Lan, Akshara Prabhakar, Silvio Savarese, Huan Wang, Shelby Heinecke
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02798v1

Many decision-support settings require systems that adapt to individual users, but evaluation data for this problem remain limited. Existing benchmarks for user understanding often rely on simulated users or model-generated behavior, even though recent work cautions that model-based simulations can diverge systematically from human behavior. We introduce \textsc{BehaviorBench}, a benchmark for evaluating personalized decision modeling from real-world behavioral traces. \textsc{BehaviorBench} reconstructs wallet-level decision histories from observed public prediction-market and on-chain records, and organizes them into two complementary task layers: \emph{Belief prediction}, which predicts a user's final revealed stance and confidence in a market, and \emph{Trade prediction}, which predicts the direction and amount of individual transactions. Across 2,000 evaluation wallets, the benchmark contains 141,445 Belief instances and 1,485,972 Trade instances, with disjoint support pools for retrieval-based evaluation. We evaluate frontier and open-weight generative models under four history interfaces: no personalization, direct recent history, generated user profiles, and retrieved support-wallet evidence. Personalization improves Belief prediction more consistently than Trade prediction, model rankings change across task layers and metrics, and different history interfaces expose different failure modes. \textsc{BehaviorBench} provides an evaluation setting for studying whether personalized methods can use real-world behavioral evidence rather than simulated users alone.

---

### Recovering Direct Price Effects of Environmental Amenities in Housing Markets: Regression and Causal Machine Learning Model Assessment with Empirical Monte Carlo Simulation

Authors: Zhenshan Chen, Klaus Moeltner, Matthew Mair
Published: 2026-06-01
Categories: econ.EM, stat.ML
arXiv: https://arxiv.org/abs/2606.02795v1

Hedonic price models are widely used to assess how environmental amenities affect property values, yet methodological guidance for estimating direct price effects remains sparse. We conduct an empirical Monte Carlo simulation to evaluate the performance of traditional and causal machine learning approaches for estimating the direct unmediated price effect of spatially delineated amenities on treated properties (DUET), a conservative lower-bound approximation for welfare changes with direct applications to benefit-cost analysis. Where previous simulations rely on parametric assumptions, we retain the actual data-generating process underlying over 1 million property transactions from upstate New York (1990--2024). By randomly assigning "treatment locations" across iterations we establish a "ground truth" that allows us to precisely measure estimation error. Our results demonstrate that generalized difference-in-differences (DID) regression consistently outperforms baseline DID and two-way fixed effects models across all scenarios. Causal Machine Learning (CML) methods, particularly causal forest DID, achieve comparable performance to generalized DID in most scenarios. In larger samples (above 3,000 treated) increasingly common in contemporary hedonic studies, CML approaches offer substantial advantages when properly specified. Based on empirical simulation results, we provide a set of method-specific best practice recommendations for both traditional regression and causal machine learning approaches.

---

### Evaluating Transformer and LSTM Frameworks for Prediction in Ungauged Basins

Authors: Taye Akinrele, James Halgren, Noorbakhsh Amiri Golilarz, Sudip Mittal, Shahram Rahimi
Published: 2026-06-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2606.02791v1

Watershed networks exhibit convergent topologies in which multiple tributaries merge into downstream channels,integrating diverse upstream hydrological processes. In ungauged basins, the absence of direct observations increases uncertainty and limits the ability to anticipate extreme events. This study evaluates whether an encoder-only Transformer provides an advantage over an LSTM for upstream streamflow inference under limited hydrologic information, using retrospective simulations from the NOAA National Water Model (NWM). Across both upstream-only and combined configurations, the LSTM showed stronger overall performance than the Transformer model across the two configurations. Incorporating downstream information further boosted performance for all models, increasing median NNSE by more than 60%. Rather than treating this as a leaderboard-style comparison, we interpret the experiments as a test of architectural inductive bias for hydrologic sequence inference. The results indicate that recurrent memory remains better aligned with this upstream reconstruction task than an encoder-only Transformer, while downstream hydrologic context provides a strong auxiliary constraint that substantially improves prediction skill across architectures

---

### Diagnosis of Human Object Interaction Detectors for Real World Educational Applications

Authors: Divya Mereddy, Ashwin Tudur Sadashiva, Marcos Quinones-Grueiro, Gautam Biswas
Published: 2026-06-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2606.02789v1

Human-object interaction (HOI) recognition is critical for automatically analyzing student behavior in complex educational environments. Although state-of-the-art (SOTA) HOI detectors perform well on benchmark datasets, their performance often degrades when deployed in real-world training environments due to domain-specific objects, occlusions, and complex visual conditions. In this paper, we introduce a diagnosis-driven framework that integrates a triplet-level HOI error taxonomy with error-factor attribution analysis for real-world educational video data. We study this problem in the context of Critical Care Air Transport Team (CCATT) mixed-reality medical training. Based on an analysis of HOI failure modes and their causes, we develop a diagnosis-informed refinement strategy for adapting pretrained HOI models to the target domain. Experiments on the CCATT dataset show that this approach improves the macro-F1 score of a pretrained CDN model from 48.6 to 90.2 through targeted refinement guided by diagnosed error factors. These results highlight the value of detailed diagnostic analysis for informing targeted adaptation of HOI models in real-world educational environments.

---

### Neutrino Fingerprints: Image-Based Encodings of IceCube Events for CNN Direction Reconstruction

Authors: Floriano Tori, Brecht Verbeken, Vincent Ginis
Published: 2026-06-01
Categories: astro-ph.IM, cs.LG
arXiv: https://arxiv.org/abs/2606.02788v1

Reconstructing the direction of incoming neutrinos in the IceCube Neutrino Observatory is an important problem in astrophysics. The public IceCube--Neutrinos in Deep Ice Kaggle competition provided 140 million simulated events to benchmark reconstruction techniques. To address this challenge from a novel perspective we introduce neutrino fingerprints compact $72 \times 72 \times 3$ images in which each pixel represents a single detector, with pulse timing and charge statistics encoded as color channels. This representation transforms sparse, irregular pulse data into dense images suitable for convolutional processing. Our ResNet18 model achieves a mean angular error of $1.10$ rad, indicating that convolutional networks trained on fingerprints rival more complex architectures while offering an effective, interpretable baseline for IceCube event reconstruction.

---

### QUIVER: Quantum-Informed Views for Enhanced Representations in Large ML Models

Authors: Aritra Bal, Michael Binder, Markus Klute, Benedikt Maier, Michael Spannowsky
Published: 2026-06-01
Categories: cs.LG, hep-ex, physics.atom-ph, quant-ph
arXiv: https://arxiv.org/abs/2606.02785v1

Large machine learning models benefit substantially from multimodal inputs that provide a complementary view of the same example. We introduce QUIVER (QUantum-Informed Views for Enhanced Representations, a paradigm that enriches classical data-driven features with a quantum Fisher view: a geometrically motivated, basis-independent summary of higher-order correlations captured by a variational quantum circuit (VQC) trained to perform the same task. Unlike classical feature augmentation, the quantum Fisher information matrix encodes the intrinsic geometry of the learned quantum state manifold. While this feature map, motivated by quantum information theory, is ordinarily non-trivial to model classically, it can surface statistical structure that additional classical data or model capacity finds difficult to learn. This makes the quantum Fisher view a genuinely complementary modality rather than a redundant one. We demonstrate that QUIVER improves standard performance metrics on two benchmark datasets from very different fields: QM9 for predicting molecule properties, and JetClass for predicting jet flavor at the Large Hadron Collider (LHC). The core contribution, however, is domain-agnostic: the quantum Fisher view can be fused into a broad class of model architectures via targeted modifications to the base architecture, to incorporate information about the quantum geometry of the problem. These results demonstrate that quantum-geometric features, extracted from simulated variational circuits, can deliver measurable value for standard machine learning tasks, well before the advent of fault-tolerant quantum hardware.

---

### CRAM-ER: Error-Resilient Spintronic Computational Random Access Memory for Scalable In-Memory Computation

Authors: Sohan Salahuddin Mugdho, Md. Shahedul Hasan, Brahmdutta Dixit, Yang Lv, Jian-Ping Wang, Cheng Wang
Published: 2026-06-01
Categories: cs.AR, cs.AI, cs.ET
arXiv: https://arxiv.org/abs/2606.02781v1

Deep neural networks (DNNs) have achieved state-of-the-art performance across diverse domains. However, typical Von Neumann compute paradigms face severe memory bottlenecks. Emerging near-memory and compute-in-memory approaches alleviate this but incur significant peripheral overhead. Computational Random Access Memory (CRAM) based on MRAM enables in-situ logic without peripheral overhead, offering a dense, energy-efficient solution. However, probabilistic MRAM switching induces gate-level errors that limit the scalability and reliability of CRAM for accelerating DNN. Moreover, the large number of sequential MRAM writes severely constrains CRAM throughput. To address these challenges, we propose an error-resilient CRAM (CRAM-ER) architecture for scalable in-memory matrix-vector multiplications (MVMs). Our error-aware hardware-software co-design framework leverages a hybrid spintronic-CRAM + CMOS adder-tree architecture to mitigate the impact of device-level errors, demonstrating MVM functionality with high area and energy efficiency. We further develop an error-aware model fine-tuning and fine-grained error correction for enhanced error resilience. Evaluations of the CMOS+spintronic hybrid architecture on DNN benchmarks show near-lossless accuracy while reducing CRAM latency by up to 2 orders of magnitude, outperforming CPU/GPU+high-bandwidth DRAM in both energy efficiency and energy-delay product.

---

### Do Value Vectors in Deep Layers Need Context from the Residual Stream?

Authors: Muyu He, Yuchen Liu, Qingya Huang, Li Zhang
Published: 2026-06-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2606.02780v5

The success of the transformer architecture is in large part due to its use of attention layers. An attention layer follows the standard neural network paradigm: it takes the residual stream as input and thereby produces context-dependent query, key, and value vectors. However, we find that model performance meaningfully improves when deeper layers learn only a context-free value vector to preserve the original token information, without drawing on any context from the residual stream. When the model has access to this context-free value vector, adding back the context-dependent component provides little additional benefit for aggregate benchmark performance. Such context-free value vectors can be stored as sparse model parameters, eliminating the need to recompute or persistently cache these values. Through systematic ablations on the key design choices for such context-free value vectors, we propose Bank of Values (BoV), a new way of computing value vectors in attention by learning a lookup table of token-specific value vectors for each of the last third of layers. Across 135M and 780M models, BoV improves validation loss over standard attention and, at 780M, the average score across 21 benchmarks, matching the previous best method that adds token information to the value vector with less compute and memory.

---

### Do Value Vectors in Deep Layers Need Context from the Residual Stream?

Authors: Muyu He, Yuchen Liu, Qingya Huang, Li Zhang
Published: 2026-06-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2606.02780v3

The success of the transformer architecture as the backbone of modern LLMs is in large part due to its use of attention layers. An attention layer follows the standard neural network paradigm: it takes the residual stream as input and thereby produces context-dependent query, key, and value vectors. However, we find that model performance meaningfully improves when deeper layers learn only a context-free value vector to preserve the original token information, without drawing on any context from the residual stream. When the model has access to this context-free value vector, adding back the context-dependent component provides little additional benefit for aggregate benchmark performance. Such context-free value vectors can be stored as sparse model parameters, eliminating the need to recompute or persistently cache these values. Through systematic ablations on the key design choices for such context-free value vectors, we propose Bank of Values (BoV), a new way of computing value vectors in attention by learning a lookup table of token-specific value vectors for each of the last third of layers. Across 135M and 780M models, BoV improves validation loss over standard attention and, at 780M, the average score across 21 benchmarks, matching the previous best method that adds token information to the value vector with less compute and memory.

---