SERO attacks a core problem in long-horizon agentic RL: outcome-based credit assignment cannot teach behaviors that never appear in the policy's own rollouts. The idea is to restore snapshots of states actually reached by the current policy, let a training-only oracle continue interacting with the real environment, and distill the resulting executable multi-step trajectories into the learner via an auxiliary reachability-weighted loss. This extends the supervision horizon as the policy improves and remains compatible with existing group-based RL algorithms, providing dense grounded guidance without touching the base estimator.

The paper tackles the ambiguity in prompt-injection detection: internal-state probes may detect injection presence rather than actual behavioral compromise. It evaluates five adapted methods on the same structured prompts for two models, stratifying by attack success, and finds that attention- and role-based signals correlate more strongly with successful attacks than presence- or recognition-based probes.

Infectious disease surveillance is hampered by fragmented data sources, uneven evidence availability across regions, and manual processing delays. EpiGuard proposes an autonomous LLM agent that proactively monitors surveillance feeds, dynamically selects forecasting strategies across absent, sparse, and abundant evidence regimes, and accumulates reusable standard operating procedures from successful warnings. The platform promises faster, more reliable early warning for over 50 diseases, but its experimental validation covers only 15 time series and never tests the absent or sparse branches it advertises.

Recommendation systems are opaque black boxes that curate feeds for billions of users, raising serious assurance risks. xLightGCN targets the popular LightGCN architecture and turns its linear graph propagation into an exact accounting of which user and item neighbors most influence a prediction. The result is a compact, inspectable subgraph plus a quantitative fidelity score that tells auditors how much of the score that subgraph actually explains—an idea that matters because interpretability alone does not guarantee the explanation covers the model’s reasoning.

This paper tackles physical reasoning in multimodal language models by asking when chain-of-thought (CoT) suffices and when physics-grounded video simulation is necessary. The authors propose Hybrid Simulation and Reasoning (HSR), a small MLP-based router that dynamically assigns questions to either CoT or a video diffusion simulator, aiming to balance accuracy against computational cost. The core claim is that this resource-rational tradeoff improves performance while allowing a 4B-parameter model to control routing for much larger base models like Gemini 3 Pro.

Static benchmarks for LLM agents saturate and face contamination, making it hard to measure true progress over time. EvoProbe tackles this by evolving cybersecurity CTF tasks in a closed loop: generate executable tasks, diagnose solver trajectories with multi-judge scoring and phase-level subtask outcomes, and feed distilled bottlenecks back into task regeneration. Its core premise is that fine-grained diagnostics—not just aggregate pass rates—should dictate what the next benchmark probes.

Cross-lingual financial analysis requires models to both recognize shared accounting concepts across languages and exhaustively recover all corresponding numerical facts. This paper introduces MultiFinGround, a benchmark built from real FY2025 ESEF reports in English, French, and Greek, aligned through the IFRS Accounting Taxonomy. The core finding is that strong concept recognition does not imply complete evidence grounding: even top models reach near-perfect recognition accuracy while missing target facts in over a quarter of correctly recognized cases. This matters because benchmarks that only test recognition can obscure critical failures in numerical evidence completeness.

Legal reasoning is not merely the mechanical application of rules to facts; it is shaped by a judge's underlying legal philosophy. This paper introduces LePhilBench, the first benchmark to test whether LLMs can both understand jurisprudential tendencies and apply them to predict judicial outcomes. It pairs jurisprudential profiling with philosophy-aware judgment prediction and grounds the tasks in a dataset of 220 annotated U.S. Supreme Court opinions, offering a rigorous new lens on models' legal reasoning capabilities.

Multilingual speech technologies still fail to reach most of the world's languages because existing corpora are limited to a few hundred well-resourced languages and narrow domains like parliamentary proceedings or Bible readings. ELARSpeech addresses this gap by curating 30,202 hours of speech across 675 underrepresented languages from the Endangered Languages Archive, yielding both labeled and unlabeled subsets with rich acoustic and communicative diversity. It matters because it provides a large-scale, publicly available testbed that reveals existing self-supervised speech models struggle on the true linguistic long tail despite massive pretraining.

Base models are hard to evaluate on agentic tasks because standard benchmarks assume instruction-following chat interfaces. ASAP adapts existing agent scaffolds to pretrained base models by treating execution as text continuation, using a Markdown+XML format, a short reference trajectory prefix, and minimal failure recovery. The framework yields discriminative end-to-end scores across models and checkpoints, and exposes execution trajectories that enable fine-grained behavioral diagnosis before post-training.

STRATA tackles the problem of scalable, self-maintaining memory for test-time self-evolving LLM agents. Instead of a flat, monotonic pool of episode summaries, it organizes experience as a typed three-layer directed graph of atomic skills, task workflows, and meta-frameworks, maintained by five mechanisms including multi-dimensional similarity, adaptive reflection thresholds, community detection, pressure-triggered garbage collection with articulation-point protection, and contrast edges that warn against directionally opposite experiences. The authors show that this structured memory attains the best average accuracy across four interactive benchmarks while simultaneously using the lowest latency, fewest environment steps, and fewer injected memory lines than seven memory-augmented baselines.

Configuring distributed LLM training over hundreds of GPUs involves a combinatorial explosion over tensor, pipeline, expert, context, and data parallelism plus micro-batch size, recomputation, and sharding, where a single byte of activation accounting can separate a legal run from an OOM crash. This paper proposes Primus Projection, a hybrid engine that benchmarks compute kernels on as few as one GPU and analytically reconstructs out-of-scope communication and pipeline schedules at the target scale, coupled with an LLM-driven tuning agent that searches the space in minutes instead of days. The authors validate across dense and MoE models on two AMD GPU generations, reporting projection errors around $5\%$ at up to 1,024 GPUs and a $+27\%$ throughput gain discovered in under 30 minutes of single-node exploration.

ixi-sLM tackles the problem of running an AI call agent entirely on a mobile device under constraints that cloud deployments never face: roughly 1 GiB of memory, in-call thermal limits, static-integer NPU execution, Korean STT input riddled with disfluencies, and user expectations set by cloud LLMs. The core idea is to take a 1.2B-parameter Korean sLM and apply two service-specific optimizations—domain-adaptive vocabulary trimming and tokenizer merging to shrink footprint and token counts, followed by task-aware mixed-precision quantization calibrated on real calls—to squeeze the model onto Qualcomm and Apple NPUs. It matters because the authors claim a live production deployment serving several million subscribers at 94% of an unnamed cloud LLM's quality, cutting per-call energy to 23% of a CPU baseline.

LLM agents face an expanded attack surface because malicious content can hide in user queries, tool descriptions, and environment observations. AgentSafeEvo casts agent safety as an attacker-defender co-evolutionary game over multi-turn tool interactions, where the attacker constructs full adversarial tasks and the defender learns to execute tasks safely. A novel learnability-guided task weighting mechanism uses within-task reward variance to shape updates for both players. Experiments on Qwen3-8B show strong safety gains and broad preservation of general capabilities.

EvoWeave addresses how interactive agents can reuse past experience without importing entire workflows or discarding step dependencies. It proposes a graph memory where nodes are abstract states and edges are reusable operations annotated with contracts, provenance, and capability references; a task planner guides retrieval and bounded graph search to assemble experience fragments. Experiments across four benchmarks and three LLM backbones show heterogeneous gains and highlight tradeoffs between retrieval precision and runtime overhead.

Program synthesis and library learning generate useful operations, but the resulting artifacts are often treated merely as derivations or search traces rather than reusable operational objects. This paper introduces FormCraft, a framework that represents both primitive and composed operations as first-class capabilities in a uniform well-formed space $\mathfrak{F}$, separating instance-independent formation and public callability from optional persistence through a verification gate. By retaining formation provenance while exposing composites through the same public interface as primitives, the framework aims to make persistent library growth possible and to eliminate the need to reconstruct historical derivation chains at reuse time, a claim backed by a paired experiment showing orders-of-magnitude reductions in search work.

Choosing a parallelism configuration for distributed LLM inference is challenging because the prefill and decode phases impose conflicting demands on time-to-first-token (TTFT) and inter-token latency (ITL). This paper develops a systematic framework that defines parallelism operating regions via TTFT-ITL Pareto frontiers and formulates SLO-aware selection using goodput $G_p(w,T_s)=\lambda_p(w)F_p(T_s|w)$. The core finding is a workload-dependent transition: tensor-parallel (TP) configurations excel for decode-heavy workloads and relaxed latency requirements, while pipeline-parallel (PP) configurations dominate for prefill-heavy workloads, stringent TTFT targets, and long contexts. The work matters because it moves parallelism selection from a static deployment choice to a workload- and SLO-dependent operating decision.

Long-horizon language agents frequently re-issue identical tool calls and receive redundant observations, a failure mode the authors call a loop. The paper shows that compressing the agent's history to fit a context window dramatically increases these loops, which in turn strongly predict task failure. DeLoop solves this by mining preference pairs directly from the agent's own rollouts—treating the looping action as a rejected example and the first later non-looping action as a chosen one—and training with direct preference optimization without any external rewards, gold answers, or LLM judges.

KoAGE investigates whether large language models can know, learn from context, and coin Korean neologisms. The authors introduce NeoTrace, an automated pipeline that collects and dates 4,271 neologisms across six genres, and propose a cutoff-aware benchmark that separates lexical recall from contextual inference and generation. By comparing model outputs with attested human coinages under original materials and simulated community settings, the paper offers a temporally grounded view of how models adapt to evolving vocabulary.

This paper exposes a critical flaw in how compressed memory is evaluated for large language models: standard ablation tests like zeroing memory vectors do not prove that stored document content improves predictions. The authors distinguish three empirically separable questions—whether a model depends on memory after training, whether training with memory outperforms training without it, and whether gains actually require document content—and argue that evaluation claims must be matched to specific controlled comparisons. Across nine memory systems and multiple QA benchmarks, they show these comparisons frequently lead to different conclusions, calling for stricter evidentiary standards in context-compression research.