Recent studies have suggested that Large Language Models (LLMs) could provide interesting ideas contributing to mathematical discovery. This claim was motivated by reports that LLM-based genetic algorithms produced heuristics offering new insights into the online bin packing problem under uniform and Weibull distributions. In this work, we reassess this claim through a detailed analysis of the heuristics produced by LLMs, examining both their behavior and interpretability. Despite being human-readable, these heuristics remain largely opaque even to domain experts. Building on this analysis, we propose a new class of algorithms tailored to these specific bin packing instances. The derived algorithms are significantly simpler, more efficient, more interpretable, and more generalizable, suggesting that the considered instances are themselves relatively simple. We then discuss the limitations of the claim regarding LLMs' contribution to this problem, which appears to rest on the mistaken assumption that the instances had previously been studied. Our findings instead emphasize the need for rigorous validation and con-textualization when assessing the scientific value of LLM-generated outputs.

Language models trained via outcome-based reinforcement learning (RL) to reason using chain-of-thought (CoT) have shown remarkable performance. Monitoring such a model's CoT may allow us to understand its intentions and detect potential malicious behavior. However, to be effective, this requires that CoTs are legible and faithful. We study CoT legibility across 14 reasoning models, finding that RL often causes reasoning to become illegible to both humans and AI monitors, with reasoning models (except Claude) generating illegible CoTs while returning to perfectly readable final answers. We show that models use illegible reasoning to reach correct answers (accuracy dropping by 53\% when forced to use only legible portions), yet find no correlation between legibility and performance when resampling - suggesting the relationship is more nuanced. We also find that legibility degrades on harder questions. We discuss potential hypotheses for these results, including steganography, training artifacts, and vestigial tokens. These results suggest that without explicit optimization for legibility, outcome-based RL naturally produces models with increasingly opaque reasoning processes, potentially undermining monitoring approaches.

Enterprise systems are crucial for enhancing productivity and decision-making among employees and customers. Integrating LLM based systems into enterprise systems enables intelligent automation, personalized experiences, and efficient information retrieval, driving operational efficiency and strategic growth. However, developing and evaluating such systems is challenging due to the inherent complexity of enterprise environments, where data is fragmented across multiple sources and governed by sophisticated access controls. We present EnterpriseBench, a comprehensive benchmark that simulates enterprise settings, featuring 500 diverse tasks across software engineering, HR, finance, and administrative domains. Our benchmark uniquely captures key enterprise characteristics including data source fragmentation, access control hierarchies, and cross-functional workflows. Additionally, we provide a novel data generation pipeline that creates internally consistent enterprise tasks from organizational metadata. Experiments with state-of-the-art LLM agents demonstrate that even the most capable models achieve only 41.8% task completion, highlighting significant opportunities for improvement in enterprise-focused AI systems.

As large language models (LLMs) enter the medical domain, most benchmarks evaluate them on question answering or descriptive reasoning, overlooking quantitative reasoning critical to clinical decision-making. Existing datasets like MedCalc-Bench cover few calculation tasks and fail to reflect real-world computational scenarios. We introduce MedCalc-Eval, the largest benchmark for assessing LLMs' medical calculation abilities, comprising 700+ tasks across two types: equation-based (e.g., Cockcroft-Gault, BMI, BSA) and rule-based scoring systems (e.g., Apgar, Glasgow Coma Scale). These tasks span diverse specialties including internal medicine, surgery, pediatrics, and cardiology, offering a broader and more challenging evaluation setting. To improve performance, we further develop MedCalc-Env, a reinforcement learning environment built on the InternBootcamp framework, enabling multi-step clinical reasoning and planning. Fine-tuning a Qwen2.5-32B model within this environment achieves state-of-the-art results on MedCalc-Eval, with notable gains in numerical sensitivity, formula selection, and reasoning robustness. Remaining challenges include unit conversion, multi-condition logic, and contextual understanding. Code and datasets are available at https://github.com/maokangkun/MedCalc-Eval.

In medical imaging, vision-language models face a critical duality: pretrained networks offer broad robustness but lack subtle, modality-specific characteristics, while fine-tuned expert models achieve high in-distribution accuracy yet falter under modality shift. Existing model-merging techniques, designed for natural-image benchmarks, are simple and efficient but fail to deliver consistent gains across diverse medical modalities; their static interpolation limits reliability in varied clinical tasks. To address this, we introduce Test-Time Task adaptive merging (T^3), a backpropagation-free framework that computes per-sample interpolation coefficients via the Jensen-Shannon divergence between the two models' output distributions. T^3 dynamically preserves local precision when models agree and defers to generalist robustness under drift. To overcome the inference costs of sample-wise merging, we further propose a batch-wise extension, T^3_B, that computes a merging coefficient across a batch of samples, dramatically reducing computational bottleneck. Recognizing the lack of a standardized medical-merging benchmark, we present a rigorous cross-evaluation protocol spanning in-domain, base-to-novel, and corruptions across four modalities. Empirically, T^3 sets new state-of-the-art in Top-1 accuracy and error reduction, outperforming strong baselines while maintaining efficiency, paving the way for adaptive MVLM deployment in clinical settings. Our code is available at https://github.com/Razaimam45/TCube.

Vision-Language Models (VLMs) excel in diverse multimodal tasks. However, user requirements vary across scenarios, which can be categorized into fast response, high-quality output, and low energy consumption. Relying solely on large models deployed in the cloud for all queries often leads to high latency and energy cost, while small models deployed on edge devices are capable of handling simpler tasks with low latency and energy cost. To fully leverage the strengths of both large and small models, we propose ECVL-ROUTER, the first scenario-aware routing framework for VLMs. Our approach introduces a new routing strategy and evaluation metrics that dynamically select the appropriate model for each query based on user requirements, maximizing overall utility. We also construct a multimodal response-quality dataset tailored for router training and validate the approach through extensive experiments. Results show that our approach successfully routes over 80% of queries to the small model while incurring less than 10% drop in problem solving probability. Vision-Language Models (VLMs), which integrate visual and textual understanding, have become crucial components in a wide range of AI applications, from robotics control to user interface navigation (Zhang et al., 2024; Shinde et al., 2025; Li et al., 2024). Moreover, a one-size-fits-all deployment strategy is suboptimal, as users increasingly expect systems that not only deliver high-quality responses but also adapt to diverse real-world scenarios with varying demands for latency, cost, and privacy. To effectively integrate the strengths of both L VLMs and SVLMs, edge-cloud collaborative routing(Y uan et al., 2025; Hao et al., 2024) is a natural fits. At its core is a lightweight model router (Ding et al., 2024; Ong et al., 2024) that inspects each query and selects an appropriate VLM. However, a general router is insufficient, routing must be scenario-aware: behaviors vary across diverse application contexts and can be configured by users or automatically inferred by scenario detection algorithms (Fifty et al., 2023; Someki et al., 2025). Existing routers are often text-centric and optimize a fixed trade-off between cost and quality, failing to adapt to multimodal, scenario-aware user needs. For example, real-time games interaction prioritizes low latency, medical diagnostics emphasizes answer quality, and mobile assistants require low energy use and strong privacy (Asgari et al., 2025).

Instruction-tuned Large Language Models (LLMs) underperform on low resource, non-Latin scripts due to tokenizer fragmentation and weak cross-lingual coupling. We present LLINK (Latent Language Injection for Non-English Knowledge), a compute efficient language-as-modality method that conditions an instruction-tuned decoder without changing the tokenizer or retraining the decoder. First, we align sentence embeddings from a frozen multilingual encoder to the decoder's latent embedding space at a reserved position via a lightweight contrastive projector. Second, the vector is expanded into K soft slots and trained with minimal adapters so the frozen decoder consumes the signal. LLINK substantially improves bilingual retrieval and achieves 81.3% preference over the base model and 63.6% over direct fine-tuning in LLM-judged Q&A evaluations. We further find that improvements can be attributed to reduced tokenization inflation and a stronger cross lingual alignment, despite the model having residual weaknesses in numeric fidelity. Treating low resource languages as a modality offers a practical path to stronger cross-lingual alignment in lightweight LLMs.

Evaluating the abilities of large language models (LLMs) for tasks that require long-term memory and thus long-context reasoning, for example in conversational settings, is hampered by the existing benchmarks, which often lack narrative coherence, cover narrow domains, and only test simple recall-oriented tasks. This paper introduces a comprehensive solution to these challenges. First, we present a novel framework for automatically generating long (up to 10M tokens), coherent, and topically diverse conversations, accompanied by probing questions targeting a wide range of memory abilities. From this, we construct BEAM, a new benchmark comprising 100 conversations and 2,000 validated questions. Second, to enhance model performance, we propose LIGHT-a framework inspired by human cognition that equips LLMs with three complementary memory systems: a long-term episodic memory, a short-term working memory, and a scratchpad for accumulating salient facts. Our experiments on BEAM reveal that even LLMs with 1M token context windows (with and without retrieval-augmentation) struggle as dialogues lengthen. In contrast, LIGHT consistently improves performance across various models, achieving an average improvement of 3.5%-12.69% over the strongest baselines, depending on the backbone LLM. An ablation study further confirms the contribution of each memory component.

Recent theoretical advancement of information density in natural language has brought the following question on desk: To what degree does natural language exhibit periodicity pattern in its encoded information? We address this question by introducing a new method called AutoPeriod of Surprisal (APS). APS adopts a canonical periodicity detection algorithm and is able to identify any significant periods that exist in the surprisal sequence of a single document. By applying the algorithm to a set of corpora, we have obtained the following interesting results: Firstly, a considerable proportion of human language demonstrates a strong pattern of periodicity in information; Secondly, new periods that are outside the distributions of typical structural units in text (e.g., sentence boundaries, elementary discourse units, etc.) are found and further confirmed via harmonic regression modeling. We conclude that the periodicity of information in language is a joint outcome from both structured factors and other driving factors that take effect at longer distances. The advantages of our periodicity detection method and its potentials in LLM-generation detection are further discussed.

While Multimodal Large Language Models (MLLMs) have advanced GUI navigation agents, current approaches face limitations in cross-domain generalization and effective history utilization. We present a reasoning-enhanced framework that systematically integrates structured reasoning, action prediction, and history summarization. The structured reasoning component generates coherent Chain-of-Thought analyses combining progress estimation and decision reasoning, which inform both immediate action predictions and compact history summaries for future steps. Based on this framework, we train a GUI agent, \textbf{GUI-Rise}, through supervised fine-tuning on pseudo-labeled trajectories and reinforcement learning with Group Relative Policy Optimization (GRPO). This framework employs specialized rewards, including a history-aware objective, directly linking summary quality to subsequent action performance. Comprehensive evaluations on standard benchmarks demonstrate state-of-the-art results under identical training data conditions, with particularly strong performance in out-of-domain scenarios. These findings validate our framework's ability to maintain robust reasoning and generalization across diverse GUI navigation tasks. Code is available at https://leon022.github.io/GUI-Rise.