651 papers in the last 30 days, 12 in the 30 days before (+5325%).
Agentic Global Optimization of Site-Disordered Crystalline Materials (opens in a new tab)
arXiv - Atmospheric physics + Materials
Site-disordered crystals, in which several elements partially occupy the same atomic site, cannot be simulated atomistically without first resolving an explicit ordering, and the number of possible orderings grows combinatorially with supercell size. Because each candidate requires an expensive structural relaxation, locating the global-minimum ordering is a costly combinatorial search. Here, we p
MEMIG: Self-evolving Memory Construction for Antibody Optimization (opens in a new tab)
arXiv - AI / ML / NLP
Memory has become central to self-evolving LLM agents, which store and retrieve experience from past rollouts to guide future decisions. Existing systems represent this experience as free-form natural language and maintain it by rewriting, which fits tasks whose experience is narratable. However, in scientific discovery and optimization, experience is often a quantity measured by an oracle under a
Beyond Plausible Prediction: Safety Sufficiency in Web World Models (opens in a new tab)
arXiv - AI / ML / NLP
Web world models increasingly predict the consequences of candidate actions before agents act, yet a plausible predicted future does not necessarily support the correct safety decision. In a matched human evaluation of 3,000 predictions from ten frozen open models, 86.1% are judged semantically plausible. However, several models exhibit highly asymmetric authorisation behaviour, combining low fals
Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention (opens in a new tab)
arXiv - AI / ML / NLP
Agentic workflows are increasingly shifting towards long-horizon tasks, such as open-ended scientific work, where models benefit from long context windows. However, at ultra-long context (above 1M tokens), full attention is prohibitively expensive. A promising approach is sparse attention, in which a dedicated indexer selects a subset of keys for the attention. Yet current methods of indexing eith
SecPO: Principled Adversarial Training for Prompt Injection Security (opens in a new tab)
arXiv - AI / ML / NLP
Recognized as a top threat to AI agents, prompt injection allows instructions from untrusted data (email, documents, websites) to manipulate the agent's operation. The existing best-defended LLM, Meta-SecAlign, is trained with the DPO loss but still suffers from attack success rates (ASRs) approaching 100% under adaptive prompt injections. One reason is that DPO anchors the model around the undefe
arXiv - AI / ML / NLP
Long-horizon LLM agents play an important role in solving complex and sustained tasks, with ever-growing interaction contexts. However, these accumulated contexts will raise substantial computational cost and obscure key decision-relevant evidence. Existing methods typically train external compressors or context-management agents to rewrite context as shorter hard-token sequences, but such externa
arXiv - AI / ML / NLP
Detectors can flag hijacked large language model (LLM) agents from the traces of the tools they call. In the eleven evaluations we audited, attacks or anomalies make up 16% to 86% of the test samples, no evaluation varies that share, and six report no test on data from a second benchmark or source. We ask what such evaluations hide. On the AgentDrift and AgentDojo benchmarks we test trained detect
arXiv - AI / ML / NLP
Job applicants increasingly use large language models (LLMs) to write cover letters, while employers deploy LLMs to screen them. In this all-AI setting, we ask whether the \emph{choice of model} affects hiring outcomes independently of candidate qualifications. We introduce a multi-model auditing framework pairing eleven LLM writer agents with nine API-hosted evaluator agents across ten job descri
Towards Understanding Score-Seeking in LLMs (opens in a new tab)
arXiv - AI / ML / NLP
Production large language models sometimes reward-hack, exploiting flaws in how tasks are scored rather than performing the tasks as intended, sometimes in highly sophisticated and adversarial ways. We investigate the motivations driving this behavior in GPT 5.6 Sol, GPT 5.5, and a checkpoint of OpenAI o3 without safety training ("exp-rl-cap-late") by observing model behavior in agentic environmen
arXiv - AI / ML / NLP
Self-evolving agents rewrite a persistent state from feedback, but it is largely untested whether this lets them learn a rule nobody told them. An agent can also pass by writing down each correction as an exception, and its list then grows with every new rule. We introduce CLR-Bench, a Curriculum of Latent Rules benchmark. Each chain adds one hidden rule per level on top of the earlier ones. A rul
arXiv - AI / ML / NLP
Reinforcement learning and game theory describe the same object from two sides. Game theory says what a rational outcome of a strategic interaction is; reinforcement learning says how an agent could reach one through experience. The two literatures nevertheless keep separate vocabularies for identical structures, and the treatments that bridge them present the correspondence as analogy rather than
TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing (opens in a new tab)
arXiv - AI / ML / NLP
Natural-language skills enable large language model (LLM) agents to reuse procedures across tasks, yet most systems retrieve individual skills from a flat SkillBank without explicitly modeling the high-level strategy that governs their selection and use. We introduce TalentRL, a strategy-guided reinforcement learning framework for hierarchical routing over procedural memory. Coarse-grained talents
arXiv - AI / ML / NLP
The EU Platform Work Directive, to be transposed by December 2026, requires ride-hailing platforms to state in plain language the main parameters of their automated decisions and how a worker's behaviour affects them. Such rules are exactly what a large language model (LLM) can read, so drivers will be able to let one decide for them. We compare two main families of rebalancing policy once their r
arXiv - AI / ML / NLP
Retrieval-augmented generation is becoming agentic. A search harness writes queries, reads results, reformulates, and decides when to stop. This largely solves accuracy: a harness with only BM25/grep and MiniMax M3 answers 85% of BrowseComp-Plus and 91% on multi-hop FRAMES in our runs, without compiled or vector indexes. It exposes a context problem: an undisciplined loop repeats queries, rereads
P-ContextBench: From Personal Fact Recall to Personalized Agent Behavior (opens in a new tab)
arXiv - AI / ML / NLP
Large language models (LLMs) are increasingly used for personal tasks, yet evaluating their safe use of personal context remains difficult because such contexts are private, fragmented, dynamic, and socially situated. We introduce P-ContextBench, a realistic, verifiable, and scalable benchmark for evaluating personal-context understanding. P-ContextBench constructs population-structured, life-stor
arXiv - AI / ML / NLP
Latent world models allow reinforcement learning agents to learn policies from imagined trajectories, yet determining the optimal depth of model reliance remains a fundamental challenge. Fixed imagination horizons apply uniform prediction depths regardless of trajectory reliability, exposing target values to compounding model error or over-relying on critic approximations. We introduce density-gui
Can Personal Requirements Become Supervision for Planning? (opens in a new tab)
arXiv - AI / ML / NLP
Managing multiple chronic conditions through diet requires satisfying disease- and medication-specific requirements jointly across meals and days while preserving established eating habits. These requirements are public in clinical guidelines and drug labels, yet turning them into a feasible plan for one person remains difficult for language-model agents. In our dietitian-approved weekly planning
arXiv - AI / ML / NLP
Low-Earth orbit (LEO) satellite networks present a challenging routing environment due to rapidly changing topology, traffic, and link conditions. Existing deep reinforcement learning (DRL) routing methods typically rely on current observations and information encoded in network weights, limiting their ability to exploit successful past experiences. We propose RAMA, a retrieval-augmented multi-age
arXiv - AI / ML / NLP
rl_rob is an open-source Python library for teaching reinforcement learning. It implements the arc of a modern course, from tabular temporal-difference learning through linear function approximation to deep value-based and policy-gradient methods, with every algorithm written from first principles so that its update rule is visible in a few lines of code. All agents share one interface, so learner
arXiv - AI / ML / NLP
Recent scientific agents build knowledge recursively by retrieving evidence, constructing an intermediate state, and judging whether that state contains sufficient information. These agents are evaluated end-to-end, usually without ground truth, so the effect of each design choice is unknown. Here, we examined two design axes: (1) what the recursion builds (the state of evidence) and (2) what cont
Characterizing LLM Agent Behaviors in Module Shadowing Environments (opens in a new tab)
arXiv - AI / ML / NLP
LLM agents operating in untrusted workspaces face the risk of untrusted code execution during routine tasks, yet final reports and workspace states alone cannot reliably establish whether execution occurred. We introduce an evaluation framework that codes observation, reasoning, execution, and response behaviors from agent trajectories and aligns them with independent environmental records to reco
Counterfactual Pairwise Augmentation Improves Small Pointwise LLM Judges (opens in a new tab)
arXiv - AI / ML / NLP
Agentic systems increasingly gate individual steps with a pointwise judge that scores a response against a criterion. Small language models are attractive for this role because they can be served and specialized efficiently. Such a judge must locate the criterion's decision boundary, but pointwise labels under-determine that boundary in fine-tuning: labeled responses to different queries differ al
TraceSpec: Understanding and Improving Speculative Decoding for LLM Agents (opens in a new tab)
arXiv - AI / ML / NLP
LLM agents execute tasks through repeated model calls over evolving histories of reasoning, actions, and tool feedback. Aggregate speculative-decoding metrics merge these distinct decisions, providing limited guidance on where draft supervision is needed. We present TraceSpec, a framework connecting trajectory evaluation with adaptive draft training. It combines dynamic execution with fixed-prefix
arXiv - AI / ML / NLP
Conversational explainable AI (ConvXAI) agents often use GUS-style architectures with separate modules for intent recognition, dialogue management, and template-based response generation. While faithful to the underlying XAI methods, these pipelines are hard to adapt and often struggle with context-dependent follow-up questions. We present SimXAI, a lightweight LLM-centric conversational layer tha
RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory (opens in a new tab)
arXiv - AI / ML / NLP
Recent work shows that a frozen robot policy performs far better inside an agent harness, where control code decides when to query the policy, consult vision tools, or command the arm directly. Yet today these harnesses are built by hand and redesigned for each new robot or benchmark. We show that harness design can be automated, offering a strong alternative to finetuning on a novel setup. In Rob
arXiv - AI / ML / NLP
The use of large language models (LLMs) to coordinate multi-agent teams raises a fundamental architectural question: whether task allocation should be centralized in a single LLM controller or distributed across autonomous LLM agents, particularly when communication is unreliable. In a deliberately simplified perimeter-defense simulation against LLM-controlled intruders, we compared a central comm
arXiv - AI / ML / NLP
Given that AI agents can search the web broadly and deeply in real time, it is tempting to conclude that their outputs account for recent world events. We show that this conclusion does not hold in practice. While agents can search and easily find recent changes, the correctness of their responses depends on whether they search and what they search for. Popular agentic search benchmarks like Brows
KEEPER: Learning Control-Sufficient Textual Memory for Long-Horizon LLM Agents (opens in a new tab)
arXiv - AI / ML / NLP
Recursive textual memory can carry earlier errors into later updates, yet task performance alone does not establish memory reliability. We define self-consistency bias as the persistence of incorrect claims already present in the incoming memory. We present KEEPER, a comparative analysis of four auxiliary objectives along information-content and update-sensitivity axes within a shared recursive in
Structure-Aware Skill Retrieval and Arbitration for Optimization Modeling (opens in a new tab)
arXiv - AI / ML / NLP
Large Language Models (LLMs) can turn natural-language optimization tasks into solver-checked answers, but current agents remain brittle under narrative rewrites and often store reusable experience at the instance level. We introduce \textsc{ArchOR}, an archetype-centric skill agent that builds on \textsc{OptSkills} by adding two retrieval-focused modules. \emph{Centroid-Aware Representation Learn
Agentic Systems for Calibrating Cosmological Simulations (opens in a new tab)
arXiv - AI / ML / NLP
Calibrating a cosmological simulation is an expensive, high-dimensional parameter optimization task. Previous works employ Bayesian optimization for automation, which treats the simulator as a black box and precludes the incorporation of human domain knowledge or workflow into the selection policy. We propose AgenticTuner, an agentic calibration pipeline with a present-day large language model ser