Claims Browser

MLLM agents' local abilities do not compose into sustained goal-directed behavior over extended exploration

"Their central failure emerges over extended exploration, where local abilities do not compose into sustained goal-directed behavior and errors accumulate without effective correction"

Most existing LLM approaches to automated code review oversimplify code review into a single-round, static decision task, failing to capture the multi-round interactive nature of realistic review scenarios

"Although recent work explores large language models (LLMs) for automated code review, most approaches oversimplify code review into a single-round, static decision task, which fails to capture the multi-round interactive nature and the complex problem-solving processes inherent in realistic review scenarios"

Trajectory-based retrieval in agentic attackers can reuse misleading experiences due to retrieval bias and unclear tool credit, with full trajectories adding context overhead while reducing interpretability

"such retrieval can reuse misleading experiences due to retrieval bias and unclear tool credit, and full trajectories add context overhead while reducing interpretability"

Prior threshold-pruned beam summing methods for TLMs produce lower bounds with unknown error

"Prior work uses a computational shortcut based on source prefix probabilities, then approximates the resulting sum with threshold-pruned beam summing. This produces a lower bound with unknown error."

Pre-separation architecture analysis revealed that the governed execution path was decoupled from the persona by omission rather than by construction, creating a risk that later wiring changes could reverse isolation, which PES prevents by making it an audited architectural rule.

"A probe of a recovered pre-separation build found the governed execution path decoupled from the persona by omission, not by construction; a later wiring change could reverse that isolation, which PES makes an audited architectural rule."

Traditional inference-time scaling methods rely on repeated generation or external verification, which is a limitation

"these methods typically rely on repeated generation or external verification"

Current sepsis severity indices use outdated fixed variables and weights from decades ago that no longer reflect modern critical care practices

"Currently used sepsis severity indices rely on fixed variables and weights established decades ago, which are coarsely discretized and calibrated to a cohort that no longer reflects contemporary critical care."

Orientation and pedestrian-aware movement remain unreliable for current MLLM agents

"while orientation and pedestrian-aware movement remain unreliable"

Existing physics-integrated 3D Gaussian pipelines are limited because they assume material parameters are known or manually specified

"Existing pipelines, however, assume that material parameters are known or manually specified, limiting their applicability when these parameters must be inferred from observed object dynamics."

Current synthetic datasets for evaluating LLM document understanding have been overly simple

"synthetic datasets have been overly simple"

Recent prompt optimizer developments favor unnecessarily complex approaches

"recent developments increasingly favor unnecessarily complex prompt optimizers"

Existing propose-and-verify methods for agent harness adaptation waste rollouts on unrelated behaviors and allow aggregate scores to obscure specific regressions

"Existing propose-and-verify methods typically score every candidate on a fixed task set, wasting rollouts on unrelated behaviors and allowing aggregate scores to obscure specific regressions."

Absolute revisit scores are sensitive to rendering stability, repetitive content, and failed motion, making them unreliable metrics

"This ambiguity makes absolute revisit scores sensitive to rendering stability, repetitive content, and failed motion."

Current safety evaluation pipelines treat eval-awareness as a single quantity to be suppressed, which may be inadequate given its heterogeneous nature

"Steering interventions targeting eval-awareness, a model's recognition that it is being tested, are increasingly used in safety evaluation pipelines, where evaluation-awareness is treated as a single quantity to be suppressed"

Existing evaluations of video generation models do not test whether repeated generations recover the correct distribution of outcomes

"existing evaluations largely assess individual-video plausibility and do not test whether repeated generations recover the correct distribution"

Post-hoc chain-of-thought labels are too coarse to show how intent changes during generation

"However, post-hoc CoT labels are too coarse to show how intent changes during generation."

Current LLM-based EDA systems fall into a syntax trap where models produce plausible code rather than physically correct hardware

"this reveals a syntax trap in which models are trained to produce plausible code rather than physically correct hardware, compounded by fragmented tools and loss of design context that obscure how decisions affect later stages."

Existing visual token pruning methods fail to optimize the substantial latency of the visual encoding phase and often cannot jointly preserve holistic visual contexts and fine-grained details under strict token budgets

"Existing visual token pruning methods exhibit two fundamental limitations. First, most approaches operate exclusively post-vision encoder, leaving the substantial latency of the visual encoding phase unoptimized. Second, under strict token budgets, these methods often fail to jointly preserve holistic visual contexts and fine-grained details, leading to performance degradation."

Recent LLM-driven personal-health agents mainly handle short-term, retrieval-based lookups and do not evaluate whether agents can reason over long-term signals

"Recent LLM-driven personal-health agents enable natural language queries over wearable signals, but mainly handle short-term, retrieval-based lookups (e.g., highest step count over a week). They do not evaluate whether agents can reason over long-term signals to predict wellbeing scores paired with evidence-grounded rationales."