Academic Intelligence · Curated Daily

Explore the Frontier of Global Academia

AcademicHub aggregates real-time literature from top journals and preprint platforms. Build your personal research radar and let large language models compile cross-disciplinary analysis briefings automatically.

01.
arXiv (quant-ph) 2026-06-12

To Cool, or Not to Cool? Displacement Sensing with Hot Quantum States

arXiv:2606.13650v1 Announce Type: new Abstract: Quantum-enhanced displacement sensing with bosonic systems is typically formulated assuming that the oscillator is cooled close to its ground state before nonclassical probe preparation. We investigate whether such near-ground-state initialization is necessary, or whether sensitive probes can instead be generated directly from thermal states. We analyze hot quantum probes produced by squeezing, number-raising, and Schrödinger-cat-state generation applied to thermal inputs. We identify two distinct mechanisms by which thermal mixedness can remain compatible with enhanced displacement sensitivity. First, projecting a mixed probe onto a definite parity sector removes the usual thermal suppression of the displacement quantum Fisher information, which can then increase with initial thermal occupation. Second, coherent superpositions of opposite displacements can retain sensitivity through coherence between their displaced components, even when the underlying state is mixed. We use these two mechanisms to classify hot-state protocols according to whether their sensitivity comes from parity selection, coherence between displaced components, or both. Finally, we formulate an experimentally relevant optimization problem comparing initial cooling with direct hot-state preparation under realistic decoherence and show that complete cooling is not universally optimal. Our results establish hot-state engineering as a route to quantum-enhanced bosonic displacement sensing without mandatory ground-state initialization.

02.
arXiv (CS.CL) 2026-06-16

REFLEX: Reflective Evolution from LLM Experience

Authors:

Large multimodal language models (LLMs) have emerged as powerful tools for guiding evolutionary search toward interpretable programmatic policies. However, existing frameworks rely on a monolithic model call to simultaneously interpret visual behavioral evidence and synthesize corrective code. This diagnosis-repair entanglement creates an opaque feedback loop, obscuring the rationale behind mutations and preventing the retention of algorithmic insights across independent runs. To achieve auditable and efficient policy search, we argue that visual diagnosis must be structurally decoupled from code generation. We present REFLEX, a train-free evolutionary framework that operationalizes this decoupling. In REFLEX, a vision-enabled Critic first distills task-specific behavioral evidence into structured, auditable diagnoses. Subsequently, a text-optimized Actor synthesizes child policies using these diagnoses alongside a persistent, self-evolving Skill Memory of reusable code snippets. This architecture not only provides transparent mutation traces but also enables cross-run programmatic knowledge transfer. Extensive evaluations across control benchmarks (Lunar Lander, Acrobot, Pendulum) and a 36-dimensional antenna array synthesis task demonstrate exceptional sample efficiency. Notably, REFLEX solves Acrobot and Pendulum in under 10 LLM calls and reaches a best Normalized Weighted Score of 1.092 on Lunar Lander, achieving highly competitive final performance while significantly accelerating the early-stage discovery of transparent policies.

03.
arXiv (CS.AI) 2026-06-19

LoRDO: Distributed Low-Rank Optimization with Infrequent Communication

arXiv:2602.04396v2 Announce Type: replace-cross Abstract: Distributed training of foundation models via $\texttt{DDP}$ is limited by interconnect bandwidth. While infrequent communication strategies reduce synchronization frequency, they remain bottlenecked by the memory and communication requirements of optimizer states. Low-rank optimizers can alleviate these constraints; however, in the local-update regime, workers lack access to the full-batch gradients required to compute low-rank projections, which degrades performance. We propose $\texttt{LoRDO}$, a principled framework unifying low-rank optimization with infrequent synchronization. We first demonstrate that, while global projections based on pseudo-gradients are theoretically superior, they permanently restrict the optimization trajectory to a low-rank subspace. To restore subspace exploration, we introduce a full-rank quasi-hyperbolic update. $\texttt{LoRDO}$ achieves near-parity with low-rank $\texttt{DDP}$ in language modeling and downstream tasks at model scales of $125$M–$720$M, while reducing communication by $\approx 10 \times$. Finally, we show that $\texttt{LoRDO}$ improves performance even more in very low-memory settings with small rank/batch size.

04.
arXiv (CS.CL) 2026-06-12

On Sequence-to-Sequence Models for Automated Log Parsing

Context: Log parsing is a critical standard operating procedure in software systems, enabling monitoring, anomaly detection, and failure diagnosis. However, automated log parsing remains challenging due to heterogeneous log formats, distribution shifts between training and deployment data, and the brittleness of rule-based approaches. Objectives: This study aims to systematically evaluate how sequence modelling architecture, representation choice, sequence length, and training data availability influence automated log parsing performance and computational cost. Methods: We conduct a controlled empirical study comparing four sequence modelling architectures: Transformer, Mamba state-space, monodirectional LSTM, and bidirectional LSTM models. In total, 396 models are trained across multiple dataset configurations and evaluated using relative Levenshtein edit distance with statistical significance testing. Results: Transformer achieves the lowest mean relative edit distance (0.111), followed by Mamba (0.145), mono-LSTM (0.186), and bi-LSTM (0.265), where lower values are better. Mamba provides competitive accuracy with substantially lower computational cost. Character-level tokenization generally improves performance, sequence length has negligible practical impact on Transformer accuracy, and both Mamba and Transformer demonstrate stronger sample efficiency than recurrent models. Conclusion: Overall, Transformers reduce parsing error by 23.4%, while Mamba is a strong alternative under data or compute constraints. These results also clarify the roles of representation choice, sequence length, and sample efficiency, providing practical guidance for researchers and practitioners.

05.
arXiv (CS.LG) 2026-06-17

From Compression to Deployment: Real-Time and Energy-Efficient FastGRNN on Ultra-Constrained Microcontrollers

arXiv:2606.17249v1 Announce Type: cross Abstract: The dominant trajectory of modern machine learning has been to scale up: larger models, larger accelerators, larger memory budgets. Yet a multi-year global semiconductor supply constraint and the growing energy and carbon cost of always-online inference expose the fragility of this trajectory and motivate the opposite direction: refactoring AI and ML algorithms to fit the small, ubiquitous microcontrollers already in mass production in wearables, sensors, and edge appliances. We present an end-to-end open-source reproduction of FastGRNN, a compact gated recurrent cell, deployed on two bare-metal targets: the 8-bit Arduino (ATmega328P) and the 16-bit MSP430 (no hardware multiplier; 16 KB Flash; 512 B SRAM). Our compression pipeline combines low-rank weight factorization, iterative hard-thresholding sparsity, and per-tensor Q15 post-training quantization with explicit activation calibration. The deployed model occupies 566 bytes of weights and achieves macro F1 = 0.918 (seed 0; five-seed Q15 mean 0.853+-0.107) on the HAPT test set. It matches a PyTorch reference at 100% prediction agreement across 3,399 test windows (MCU seed 0; 99.91-100% C-equivalent across five seeds). Both platforms sustain real-time 50 Hz streaming inference (9.21 ms per sample on Arduino; 13 ms on MSP430), where a 256-entry sigmoid/tanh look-up table delivers a 30.5x speedup on the multiplier-less MSP430. Four contributions extend the original FastGRNN paper: (i) cross-platform bit-equivalent deterministic inference; (ii) characterization of recurrent warm-up latency (median 74 samples, 1.48 s; worst-case 125 samples, 2.50 s over 100 test windows); (iii) a deployable look-up-table recipe for multiplier-less embedded targets; and (iv) hardware energy characterization showing 17.7 mW active inference power,

06.
arXiv (CS.CV) 2026-06-12

Learning Task-Aware Sampling with Shared Saliency through Density-Equalizing Mappings

In image and surface-based learning tasks, convolutional features are typically extracted using receptive fields that are sampled uniformly across the entire domain. However, informative structures are rarely distributed uniformly in practice and are often concentrated in localized regions. Such phenomena are particularly common in medical imaging, where pathological changes are spatially confined. Consequently, uniform convolution allocates equal computational effort to both informative and uninformative regions, resulting in inefficient feature extraction and suboptimal utilization of model capacity. To address this issue, we propose a framework for task-adaptive sampling that dynamically redistributes computational attention according to the spatial importance of the data. Specifically, we introduce the Density-Equalizing Convolutional Neural Network (DECNN), which employs density-equalizing mappings to guide convolution through a learned density function. The density function encodes the relative importance of different regions and induces a transformation that enlarges informative areas while compressing less relevant ones. As a result, convolutional receptive fields are redistributed non-uniformly over the domain, enabling denser sampling in task-relevant regions. By coupling this importance-driven transformation with convolution, DECNN performs adaptive feature extraction that focuses computational resources on informative structures. This leads to more efficient use of model capacity, yielding a lightweight yet expressive architecture while simultaneously producing an interpretable saliency map. Experiments on image classification and craniofacial surface analysis demonstrate that DECNN achieves competitive or superior performance with fewer parameters, accurately identifies task-relevant regions, and remains robust under complex geometric variations.

07.
arXiv (CS.CV) 2026-06-12

Revisiting Vehicle Color Recognition in Long-Tailed Surveillance Scenarios

Vehicle color recognition is an important cue for vehicle identification in surveillance systems, especially when license plates are illegible due to low resolution, occlusion, motion blur, or poor illumination. However, real-world vehicle color distributions are highly imbalanced, making overall accuracy insufficient to assess performance on rare but operationally relevant colors. This paper presents a comprehensive study of vehicle color recognition under severe class imbalance using UFPR-VeSV, a challenging real-world surveillance dataset. We investigate synthetic minority-class augmentation through two off-the-shelf generative strategies: text-conditioned image generation with RunDiffusion/JuggernautXL and image-conditioned color editing with Gemini 2.0 Flash. The curated synthetic data are combined with modern visual representations, loss reweighting, learning-rate scheduling, color-safe augmentation, foreground-aware preprocessing, and ensemble fusion. The bestperforming approach achieves 94.6% micro accuracy and 79.7% macro accuracy, improving macro accuracy by 8.2 percentage points over recent literature. A manual error analysis further shows that many remaining failures are visually ambiguous even for human annotators, highlighting the practical limits of color-based vehicle identification in unconstrained surveillance imagery. The generated images and source code are publicly available at https://github.com/viniciusorru/vcr-synthetic

08.
arXiv (CS.AI) 2026-06-24

Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation

arXiv:2606.24515v1 Announce Type: new Abstract: Computer-Use Agents (CUAs) execute high-level user goals by perceiving and acting directly within graphical user interfaces. However, reinforcement learning for CUAs remains difficult because open-ended desktop environments rarely provide scalable, machine-readable reward signals: task success is often visually grounded and hard to specify with handcrafted reward functions or dense manual labels. We propose an RL fine-tuning framework that uses autonomous vision-language evaluation as a scalable supervision signal for GUI agents. Given a final screenshot and the original instruction, a Vision-Language Model judges task completion and provides terminal feedback without task-specific heuristics or manual labels during policy optimization. Because autonomous evaluators are imperfect, we model their feedback as a noisy binary reward channel and derive a noise-corrected reward estimator for Proximal Policy Optimization. Experiments across macOSWorld, Windows Agent Arena, and OSWorld show that corrected evaluator rewards outperform both zero-shot baselines and raw evaluator rewards, improving success rates by an average of 12.6 percentage points over zero-shot performance and 5.1 points over raw evaluator fine-tuning. These results suggest that autonomous evaluation can serve as a practical reward signal for RL in GUI environments when evaluator noise is explicitly modeled and corrected.

09.
medRxiv (Medicine) 2026-06-22

Associations of Chemical Exposures with Psychological Distress and Depression Diagnosis among Waste Pickers in Brasilia, Brazil: A Cross-Sectional Study

Introduction: Waste pickers face chemical exposures. We evaluated whether chemical exposure is associated with psychological distress and depression. Methods: A 2017 cross-sectional survey included 1,141 waste pickers working in the Estrutural open dump in Brasilia, Brazil. Participants self-reported occupational exposure to 11 chemical categories, 17 psychological distress symptoms, and depression diagnoses. Associations of chemical exposure with mean psychological distress scores and depression prevalence were assessed, adjusted for age, sex, marital status, and income. Results: Mean psychological distress score was higher among those exposed to any chemical (mean of 8.1 vs 6.1; adjusted mean difference [aMD]: 1.8 [0.9, 2.7]) and higher among those exposed to each of 11 chemical categories, for example, smoke (aMD: 1.2 [0.6, 1.7]), batteries (aMD: 1.5 [1.0, 1.9], and oils (aMD: 1.3 [0.9, 1.8]). Depression was more prevalent among those exposed to oils (16.6% vs 10.6%; adjusted prevalence difference [aPD]: 6.3% [95% CI: 2.3, 10.2]), cleaning products (aPD: 5.4% [1.2, 9.5]), medications (aPD: 4.7% [0.6, 8.8]), and aerosols (aPD: 5.3% [1.3, 9.3]) but, not smoke, batteries, greases, insecticides, solvents, paints, chemical containers, or any chemical. Conclusion: These associations highlight the need to consider policy level protections for waste pickers to reduce chemical exposure and guard against psychological distress. Further research is necessary to explore which specific chemicals, within broad chemical categories, are associated with psychological distress and depression.

10.
arXiv (CS.AI) 2026-06-16

Entropy-Gated Latent Recursion

arXiv:2606.16620v1 Announce Type: cross Abstract: Inference-time scaling has become the dominant lever for improving language-model reasoning, but existing methods derive rollout diversity from a single source: stochastic token-level sampling. We argue that this single-axis sampling space is fundamentally limiting, and identify a second, fully deterministic and complementary axis: the layer span $L$ at which a frozen model's top decoder layers are recursively re-applied at high-uncertainty tokens. Different choices of $L$ produce distinct rollouts that solve different subsets of problems, with no stochasticity. We instantiate this axis through Entropy-Gated Latent Recursion (EGLR), a training-free decoding procedure that re-applies the top-$L$ layers for at most $K_{\max}$ iterations until the next-token distribution converges. Combined with $T$ temperature samples, EGLR turns a single-axis stochastic rollout pool into an $L\times T$ Cartesian sampling space at almost the same per-rollout cost. We characterize this space across $8$ instruction-tuned models and $6$ math reasoning benchmarks, and show that the $L$-axis is genuinely complementary to temperature: on MATH-500 with Qwen2.5-3B-Instruct, the joint $L\times T$ oracle reaches $91.6\%$, $+8.2$ percentage points beyond the temperature-only oracle ($83.4\%$) and $+10.4$ points beyond the layer-only oracle ($81.2\%$), confirming that the two axes capture genuinely complementary problems. The expanded rollout pool provides richer per-prompt candidates for any downstream procedure that consumes rollouts, including self-consistency, best-of-$N$ with verifiers, and group-relative RL training (GRPO), opening a new direction for inference-time scaling that does not rely on stochastic noise.

11.
arXiv (math.PR) 2026-06-17

Killed resolvents and measure-valued stopping gains for reflected optimal stopping with max-type rewards

arXiv:2606.17517v1 Announce Type: new Abstract: We study an infinite-horizon optimal stopping problem for a normally reflected two-dimensional diffusion in the positive quadrant with nonsmooth max-type reward \(G(x_1,x_2)=x_1\vee \alpha x_2\). The paper develops a conditional measure-theoretic framework for the associated reflected obstacle problem. The main innovation is to show that the stopping gain \(\Gamma=c+rG-\mathcal LG\) is a signed measure, not a function: the kink of \(G\) generates an explicit negative surface measure on \(\Delta=\{x_1=\alpha x_2\}\). We then prove that the correct potential representation uses the resolvent of the reflected diffusion killed on first entry into the stopping set, rather than the unrestricted reflected resolvent. Under explicit monotonicity, regularity, and measure-superharmonicity assumptions, we derive an epigraph representation, a continuation-side boundary-trace condition, and a candidate verification theorem. The framework clarifies hidden regularity and uniqueness assumptions in multidimensional nonsmooth optimal stopping.

12.
arXiv (CS.CL) 2026-06-16

Misinformation Propagation in Benign Multi-Agent Systems

Multi-agent systems, in which multiple large language model agents solve problems through turn-based interaction, are increasingly deployed in high-stakes settings such as medical diagnosis, legal analysis, and forensic decision-making. Their reliability can be at risk when single agents reason from incorrect or misleading context, e.g., from tool calls, since errors may propagate through agent interactions. This work studies this risk by injecting intent-based misinformation into benign single-agent and multi-agent systems across reasoning, knowledge, and alignment tasks. We find that misinformation can degrade single-agent performance and persists across multi-agent debate, with agents often retaining answers introduced by misinformed peers. Nevertheless, multi-agent debate reduces the resulting performance degradation compared to single-agent prompting, especially when most agents are not exposed to misinformation. Robustness depends on group composition and decision protocol. Consensus can be more stable than voting under peer pressure, while majorities can often steer misinformed agents back toward correct answers. Our results show that misinformation robustness in multi-agent systems depends on the underlying model and also on how agents exchange information and aggregate decisions.

13.
arXiv (CS.AI) 2026-06-24

Grad Detect: Gradient-Based Hallucination Detection in LLMs

arXiv:2606.24790v1 Announce Type: cross Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across diverse tasks, yet they remain prone to generating hallucinations. Detecting these hallucinations is critical for deploying LLMs reliably in high-stakes applications. We present Grad Detect, a gradient-based approach for predicting hallucinations by analyzing layer-wise gradient patterns from a single forward-backward pass during inference. Our method shows that the internal gradient structure of a model carries rich information about the correctness of its output. This information is not accessible through output-level signals alone. We evaluate Grad Detect on several Q&A benchmarks across both hallucination detection and model abstention prediction, where it consistently outperforms confidence-based and sampling-based baselines. Through comprehensive layer ablation studies across all eleven models from four architectural families, we find that the final five layers concentrate over 97% of the discriminative gradient signal, enabling efficient deployment with minimal performance loss. Grad Detect provides a unified framework for predicting multiple dimensions of LLM reliability, offering strong predictive performance alongside interpretable insights into where and how model failures originate.

14.
arXiv (CS.CL) 2026-06-11

Agent Skill Evaluation and Evolution: Frameworks and Benchmarks

The growth of agent skills has transformed how agentic systems are built, evaluated, and deployed. As skill libraries continue to scale, rigorous evaluation becomes critical to ensuring their utility, quality, and safety in real-world applications. Consequently, the field is undergoing an emerging paradigm shift from isolated skill creation to automated, evaluation-driven skill evolution. In this survey, we systematically examine the landscape of skill evolution and evaluation beyond foundational skill creation. We categorize evolution into four distinct paradigms, spanning execution feedback, trajectory distillation, compression, and reinforcement learning, showing how each element contributes to improving skill utility and reliability. We also provide an analysis of six skill-centric benchmark categories, identifying structural gaps in benchmark coverage, trade-offs, and metric richness to advance skill research. Finally, we identify open directions for building skill ecosystems that are generalizable, efficient, and verifiably safe. The project URL is https://github.com/Cassie07/AgentSkill_Survey

15.
arXiv (CS.CL) 2026-06-17

Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients

Knowledge distillation transfers a teacher's competence to a small student but is brittle in the small-student regime: forcing the student to imitate logits from a much larger teacher concentrates it on the teacher's sharpest modes, hurting generalization on benchmark families beyond the training corpus. Reinforcement learning (RL) avoids logit imitation by training on the student's own rollouts. However, on questions where every rollout fails-yielding zero advantage and being silently discarded-injecting a stronger teacher's response into the policy gradient breaks the on-policy assumption and induces drift. We introduce Zone of Proximal Policy Optimization (ZPPO), inspired by Vygotsky's zone of proximal development, which keeps the teacher inside the prompt rather than the policy gradient. On hard questions, ZPPO constructs two reformulated prompts: a Binary Candidate-included Question (BCQ) pairs one correct teacher response with one incorrect student response as anonymized candidates the student must discriminate, and a Negative Candidate-included Question (NCQ) aggregates the student's wrong rollouts into a single prompt to surface their shared failure modes. A prompt replay buffer recirculates each hard question until it either graduates-the student's mean rollout accuracy on it reaches half- or is FIFO-evicted under finite capacity, amplifying BCQ and NCQ inside the student's current zone of proximal development. On the Qwen3.5 family at four student scales (0.8B-9B) with a 27B teacher, post-trained as vision-language models and evaluated on a 31-benchmark suite (16 VLM, 10 LLM, 5 Video), ZPPO outperforms off/on-policy distillation and GRPO, with the largest gains at the smallest scale.

16.
arXiv (quant-ph) 2026-06-19

Quantum Kernels are Spectral Tensor Networks

arXiv:2606.20402v1 Announce Type: new Abstract: Quantum kernels admit Fourier representations whose frequencies are determined by the data-encoding gates of the underlying feature map. We show that entangling tensor kernels are matrix product operator factorizations of the corresponding Fourier coefficient tensors, thereby identifying quantum kernels as spectral tensor networks. By grouping gate-level frequency configurations that yield the same feature-wise frequency, we obtain a grouped Fourier form that induces a more compact spectral tensor network representation of the kernel. We further show that kernel target alignment serves as a bridge between the Fourier and tensor network views. On a grid that resolves the accessible Fourier modes, it becomes the Frobenius cosine similarity between Fourier coefficient tensors. Our numerical experiments show that layered quantum kernels admit accurate representations with small bond dimension, revealing a compressibility governed by correlations between Fourier modes. This compressibility provides a diagnostic of classical representability and of whether kernel evaluation is likely to remain classically tractable.

17.
arXiv (CS.AI) 2026-06-15

Formalizing Numerical Analysis: An Agent Pipeline and Quality Audit Beyond Kernel Acceptance

arXiv:2606.14000v1 Announce Type: new Abstract: Recent work has demonstrated that coding agents can formalize entire advanced mathematics textbooks in Lean 4, yet existing efforts concentrate on branches of mathematics already well-represented in mathlib and measure success solely through kernel acceptance. We address both limitations by applying a coding agent to formalize Numerical Methods for Ordinary Differential Equations, a textbook in numerical analysis that is largely absent from mathlib, stressing the agent's capacity to develop new theory from scratch. We further introduce a systematic, reproducible three-dimensional framework for evaluating the quality of agent-produced formalizations beyond compilation: semantic correctness, Mathlib reuse, and cross-file reuse via LLM-as-judge methods. Applying this framework to our own formalization and to the released outputs of RepoProver and M2F, we uncover recurring unfaithful formalization patterns, including incomplete multi-part statements, added weakening hypotheses, and parameter restrictions, that kernel acceptance entirely obscures. Our results suggest that compilation-based metrics substantially overstate formalization quality, and we provide a reproducible audit methodology to support more rigorous evaluation of future autoformalization systems.

18.
arXiv (quant-ph) 2026-06-17

A Quantum Approach to Stochastic Optimization in Insurance Underwriting

arXiv:2605.01169v2 Announce Type: replace Abstract: The presence of stochastic elements in combinatorial optimization problems makes them particularly challenging, as such problems quickly become intractable for classical computers even at relatively small sizes. In this work, we propose a novel quantum-classical hybrid scheme for solving a class of stochastic optimization problems known as chance-constrained knapsack problems, in which item weights follow probability distributions and constraints may be violated within a specified risk tolerance. Our method employs knapsack-specific QAOA-based circuits to generate samples which, when combined with a new self-consistent classical recovery scheme introduced in this work, produce high-quality solutions. Experiments carried out on IBM Heron processors, using circuits with depths up to 177 and comprising 3443 gates acting on as many as 150 qubits, yield solutions that indicate performance comparable to classical optimization schemes. The proposed quantum-classical scheme paves the way to tackling such problems, with the potential to outperform approaches that rely solely on classical computation.

19.
arXiv (CS.LG) 2026-06-11

RePAIR: Predictive Self-Supervised Representation Learning in Chess

arXiv:2606.11860v1 Announce Type: new Abstract: In this paper, we introduce Representation Prediction via Autoencoding using Iterative Refinement (RePAIR) - a novel self-supervised representation learning architecture that synthesizes Masked Autoencoders (MAE), Joint Embedding Predictive Architectures (JEPA), and Bidirectional Encoder Representations from Transformers (BERT). We demonstrate how it can be used to encode objects in sequential data like consecutive chess positions into compact yet meaningful representations. The basic principle of the architecture is to mask large portions of a sequence of latent states, similar to BERT and MAE. Then, we apply a lightweight Predictor to the latent representations that repairs gaps in the sequence in a lower-dimensional embedding space akin to JEPA. Our experiments in the domain of chess show that the Encoder refines the board representations such that meaningful chess concepts emerge clustered in the latent space. Furthermore, reconstructions of the masked board states show that the model is able to reason about the piece movements without relying on costly reinforcement learning methods. Lastly, we find that the resulting representation space allows for quick and intuitive dissections of chess games by observing the game path trajectories in this semantically rich space.

20.
arXiv (CS.AI) 2026-06-18

What Must Generalist Agents Remember?

arXiv:2606.18746v1 Announce Type: new Abstract: This paper develops a formal account of what generalist agents must store in memory in order to act near-optimally across multiple environments and goals. It shows that when two domains share an observational bottleneck but require incompatible optimal actions, any uniformly near-optimal policy must induce distinct memory distributions at that bottleneck. The result yields a separation theorem: sufficiently successful agents cannot rely only on current state observations, but must preserve domain-relevant information in memory. The paper further shows that if an agent's memory contains enough information to estimate values for related goals, then that memory can be used to approximately reconstruct the agent's local transition dynamics. Together, these results characterize memory as the substrate that supports domain disambiguation, transition-model reconstruction, and planning for generalist agents.

21.
arXiv (CS.LG) 2026-06-18

Strategic Feature Selection

arXiv:2606.18867v1 Announce Type: new Abstract: When algorithmic predictors inform resource allocation in high-stakes domains such as healthcare, these predictors must account for strategic manipulation of input features. The typical solution is to redesign the predictor itself to explicitly account for strategic interactions. In practice, however, decision makers are often constrained to adjusting coarser levers within existing prediction pipelines. For example, healthcare organizations often select which features to exclude based on perceived manipulability, while using standard regularization procedures to shrink the coefficients of retained features. In this work, we initiate a formal study of strategic classification through feature selection and its interaction with ridge regularization. Our main finding is that excluding individual features based on their manipulability alone is generally suboptimal. We provide a fine-grained characterization of the performance of a feature subset under optimal regularization, yielding new insights for policy design. Motivated by this characterization, we develop a practical algorithm for jointly choosing the feature set and the level of ridge regularization. Through a real-world case study on a healthcare payments benchmark, we illustrate how our algorithm can guide the design of coarse policy levers in practice. Our results provide a principled, practical framework for mitigating the effects of strategic behavior in algorithmic decision-making systems.

23.
arXiv (CS.AI) 2026-06-17

Reversal Q-Learning

arXiv:2606.17551v1 Announce Type: cross Abstract: Iterative generative modeling techniques, such as flow matching, provide powerful tools to model complex behaviors for effective offline reinforcement learning (RL). In this work, we propose a new off-policy RL algorithm that trains a flow policy based on prior data. Our idea starts from the "expanded" Markov decision process (MDP) framework, which treats individual flow refinement steps as separate actions in an MDP. To enable off-policy RL within this framework, we apply two techniques: we generate virtual on-policy trajectories (by "reversing" flows) to make this framework compatible with prior data, and we apply a bias-and-variance reduction technique to mitigate the curse of horizon in off-policy RL. We call the resulting algorithm Reversal Q-learning (RQL). RQL has several advantages over previous flow-based RL methods: it does not suffer from backpropagation through time, makes better use of the learned value function, and directly trains the full, expressive flow policy. Through our experiments on 50 challenging simulated robotic tasks, we show that RQL leads to the best average offline RL performance compared to state-of-the-art flow-based offline RL algorithms.

25.
arXiv (CS.CL) 2026-06-16

Agentic Reinforcement Learning for Search Misaligns Instruction-Tuning

Agentic reinforcement learning (RL) trains large language models to use tools, but its impact on alignment is poorly understood. We study how agentic RL for search affects the alignment of instruction-tuned (IT) models. We find that RL-trained models inherit refusal reasoning by deflecting harmful requests into benign search queries, but this breaks down under a simple diagnostic trigger that elicits a search call before refusal can occur. Under this condition, RL models produce multi-step unsafe search actions and reasoning, reducing search query safety by up to 68.6% in Qwen and Llama models relative to their IT counterparts. The effect generalises across model families, scales, and RL algorithms. To understand why, we identify linear directions in the residual stream that control search query safety, and show that RL training progressively shifts search behaviour toward the harmful end of this direction. We thus propose representation-guided RL training, which adds a reward penalty based on projection toward the harmful search direction. Training on benign data alone, it restores IT-level alignment without reducing task accuracy and requires no additional training data. Together, our work provides the first framework for diagnosing, mechanistically analysing, and mitigating alignment degradation in agentic RL for search.