EMBGuard: Constructing Hazard-Aware Guardrails for Safe Planning in Embodied Agents

MLLM-powered embodied agents deployed in real-world environments encounter physical hazards. However, existing approaches lack explicit mechanisms for identifying hazards and reasoning about action-conditioned risks, leading agents to either miss risky interactions or over-identify risks. To address this, we propose EMBGuard, the first MLLM-based safety guardrail for embodied agents designed to decouple physical risk reasoning from agent policy. By evaluating a (visual observation, action) pair, EMBGuard identifies hazardous configurations and provides natural language explanations of potential risks. Alongside EMBGuard, we contribute EMBHazard, a training dataset of 15.1K action-conditioned pairs, and EMBGuardTest, a benchmark of 329 manually curated real-world scenarios spanning seven physical risk categories. Through compositional variation of hazards and actions, we generate diverse risky and benign scenarios that agents may encounter during planning. Despite its compact size (2B, 4B), EMBGuard achieves performance competitive with proprietary MLLMs (e.g., GPT-5.1, Gemini-2.5-Pro) while significantly reducing the false-positive rates that hinder real-time deployment. We make the code, data, and models publicly available at https://github.com/dongwxxkchoi/EMBGuard

Towards Direct Evaluation of Harness Optimizers via Priority Ranking

Harness optimization enables automated agent creation by having an optimizer agent iteratively update the harness of target agents. Despite its success, current studies evaluate optimizers solely by observing target agents' performance gains. This indirect end-improvement evaluation neglects optimizers' actions at intermediate steps, which are often erroneous and hinder agent performance. Therefore, it is unclear whether harness optimization is driven by optimizers' informed update actions or simply trial-and-error. This necessitates direct evaluation of harness optimizers. However, evaluating harness optimizers directly is non-trivial and costly due to the lack of oracle harnesses. To address this, we present a simple, low-cost design to directly evaluate them, namely priority ranking. By asking harness optimizers to rank components (e.g., tools) in a given harness by their potential to improve/hinder agent performance when updated, our design quantifies optimizer ability at the step level without expensive rollouts or manual examination. More importantly, optimizers' ranking performance correlates with their ability to improve agents in actual multi-step harness optimization, establishing priority ranking as a reliable predictor of optimization ability. Priority ranking is enabled by Shor, a collection of 182 human-verified optimization scenarios spanning across domains, designs, and time stages. Codes and data can be found at https://github.com/k59118/Harness_Optimizer_Evaluation.

On Training Large Language Models for Long-Horizon Tasks: An Empirical Study of Horizon Length

Large language models (LLMs) have shown promise as interactive agents that solve tasks through extended sequences of environment interactions. While prior work has primarily focused on system-level optimizations or algorithmic improvements, the role of task horizon length in shaping training dynamics remains poorly understood. In this work, we present a systematic empirical study that examines horizon length through controlled task constructions. Specifically, we construct controlled tasks in which agents face identical decision rules and reasoning structures, but differ only in the length of action sequences required for successful completion. Our results reveal that increasing horizon length alone constitutes a training bottleneck, inducing severe training instability driven by exploration difficulties and credit assignment challenges. We demonstrate that horizon reduction is a key principle to address this limitation, stabilizing training and achieving better performance in long-horizon tasks. Moreover, we find that horizon reduction is related to stronger generalization across horizon lengths: models trained under reduced horizons generalize more effectively to longer-horizon variants at inference time, a phenomenon we refer to as horizon generalization.

PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints

We are entering an era in which individuals and organizations increasingly deploy dedicated AI agents that interact and collaborate with other agents. However, the dynamics of multi-agent collaboration under privacy constraints remain poorly understood. In this work, we present $PAC\text{-}Bench$, a benchmark for systematic evaluation of multi-agent collaboration under privacy constraints. Experiments on $PAC\text{-}Bench$ show that privacy constraints substantially degrade collaboration performance and make outcomes depend more on the initiating agent than the partner. Further analysis reveals that this degradation is driven by recurring coordination breakdowns, including early-stage privacy violations, overly conservative abstraction, and privacy-induced hallucinations. Together, our findings identify privacy-aware multi-agent collaboration as a distinct and unresolved challenge that requires new coordination mechanisms beyond existing agent capabilities.

CONDESION-BENCH: Conditional Decision-Making of Large Language Models in Compositional Action Space

Large language models have been widely explored as decision-support tools in high-stakes domains due to their contextual understanding and reasoning capabilities. However, existing decision-making benchmarks rely on two simplifying assumptions: actions are selected from a finite set of pre-defined candidates, and explicit conditions restricting action feasibility are not incorporated into the decision-making process. These assumptions fail to capture the compositional structure of real-world actions and the explicit conditions that constrain their validity. To address these limitations, we introduce CONDESION-BENCH, a benchmark designed to evaluate conditional decision-making in compositional action space. In CONDESION-BENCH, actions are defined as allocations to decision variables and are restricted by explicit conditions at the variable, contextual, and allocation levels. By employing oracle-based evaluation of both decision quality and condition adherence, we provide a more rigorous assessment of LLMs as decision-support tools.

AgenticShop: Benchmarking Agentic Product Curation for Personalized Web Shopping

The proliferation of e-commerce has made web shopping platforms key gateways for customers navigating the vast digital marketplace. Yet this rapid expansion has led to a noisy and fragmented information environment, increasing cognitive burden as shoppers explore and purchase products online. With promising potential to alleviate this challenge, agentic systems have garnered growing attention for automating user-side tasks in web shopping. Despite significant advancements, existing benchmarks fail to comprehensively evaluate how well agentic systems can curate products in open-web settings. Specifically, they have limited coverage of shopping scenarios, focusing only on simplified single-platform lookups rather than exploratory search. Moreover, they overlook personalization in evaluation, leaving unclear whether agents can adapt to diverse user preferences in realistic shopping contexts. To address this gap, we present AgenticShop, the first benchmark for evaluating agentic systems on personalized product curation in open-web environment. Crucially, our approach features realistic shopping scenarios, diverse user profiles, and a verifiable, checklist-driven personalization evaluation framework. Through extensive experiments, we demonstrate that current agentic systems remain largely insufficient, emphasizing the need for user-side systems that effectively curate tailored products across the modern web.

Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuning

Autoregressive (AR) language models generate text one token at a time, which limits their inference speed. Diffusion-based language models offer a promising alternative, as they can decode multiple tokens in parallel. However, we identify a key bottleneck in current diffusion LMs: the long decoding-window problem, where tokens generated far from the input context often become irrelevant or repetitive. Previous solutions like semi-autoregressive address this issue by splitting windows into blocks, but this sacrifices speed and bidirectionality, eliminating the main advantage of diffusion models. To overcome this, we propose Convolutional decoding (Conv), a normalization-based method that narrows the decoding window without hard segmentation, leading to better fluency and flexibility. Additionally, we introduce Rejecting Rule-based Fine-Tuning (R2FT), a post-hoc training scheme that better aligns tokens at positions far from context. Our methods achieve state-of-the-art results on open-ended generation benchmarks (e.g., AlpacaEval) among diffusion LM baselines, with significantly lower step size than previous works, demonstrating both speed and quality improvements.

Quantifying Self-Awareness of Knowledge in Large Language Models

Hallucination prediction in large language models (LLMs) is often interpreted as a sign of self-awareness. However, we argue that such performance can arise from question-side shortcuts rather than true model-side introspection. To disentangle these factors, we propose the Approximate Question-side Effect (AQE), which quantifies the contribution of question-awareness. Our analysis across multiple datasets reveals that much of the reported success stems from exploiting superficial patterns in questions. We further introduce SCAO (Semantic Compression by Answering in One word), a method that enhances the use of model-side signals. Experiments show that SCAO achieves strong and consistent performance, particularly in settings with reduced question-side cues, highlighting its effectiveness in fostering genuine self-awareness in LLMs.

Designing Memory-Augmented AR Agents for Spatiotemporal Reasoning in Personalized Task Assistance

Augmented Reality (AR) systems are increasingly integrating foundation models, such as Multimodal Large Language Models (MLLMs), to provide more context-aware and adaptive user experiences. This integration has led to the development of AR agents to support intelligent, goal-directed interactions in real-world environments. While current AR agents effectively support immediate tasks, they struggle with complex multi-step scenarios that require understanding and leveraging user's long-term experiences and preferences. This limitation stems from their inability to capture, retain, and reason over historical user interactions in spatiotemporal contexts. To address these challenges, we propose a conceptual framework for memory-augmented AR agents that can provide personalized task assistance by learning from and adapting to user-specific experiences over time. Our framework consists of four interconnected modules: (1) Perception Module for multimodal sensor processing, (2) Memory Module for persistent spatiotemporal experience storage, (3) Spatiotemporal Reasoning Module for synthesizing past and present contexts, and (4) Actuator Module for effective AR communication. We further present an implementation roadmap, a future evaluation strategy, a potential target application and use cases to demonstrate the practical applicability of our framework across diverse domains. We aim for this work to motivate future research toward developing more intelligent AR systems that can effectively bridge user's interaction history with adaptive, context-aware task assistance.

ToolHaystack: Stress-Testing Tool-Augmented Language Models in Realistic Long-Term Interactions

Large language models (LLMs) have demonstrated strong capabilities in using external tools to address user inquiries. However, most existing evaluations assume tool use in short contexts, offering limited insight into model behavior during realistic long-term interactions. To fill this gap, we introduce ToolHaystack, a benchmark for testing the tool use capabilities in long-term interactions. Each test instance in ToolHaystack includes multiple tasks execution contexts and realistic noise within a continuous conversation, enabling assessment of how well models maintain context and handle various disruptions. By applying this benchmark to 14 state-of-the-art LLMs, we find that while current models perform well in standard multi-turn settings, they often significantly struggle in ToolHaystack, highlighting critical gaps in their long-term robustness not revealed by previous tool benchmarks.