Open-source large language models (LLMs) have demonstrated considerable dominance over proprietary LLMs in resolving neural processing tasks, thanks to the collaborative and sharing nature. Although full access to source codes, model parameters, and training data lays the groundwork for transparency, we argue that such a full-access manner is vulnerable to stego attacks, and their ill-effects are not fully understood. In this paper, we conduct a systematic formalization for stego attacks on open-source LLMs by enumerating all possible threat models associated with adversary objectives, knowledge, and capabilities. Therein, the threat posed by adversaries with internal knowledge, who inject payloads and triggers during the model sharing phase, is of practical interest. We go even further and propose the first stego attack on open-source LLMs, dubbed SASER, which wields impacts through identifying targeted parameters, embedding payloads, injecting triggers, and executing payloads sequentially. Particularly, SASER enhances the attack robustness against quantization-based local deployment by de-quantizing the embedded payloads. In addition, to achieve stealthiness, SASER devises the performance-aware importance metric to identify targeted parameters with the least degradation of model performance. Extensive experiments on LlaMA2-7B and ChatGLM3-6B, without quantization, show that the stealth rate of SASER outperforms existing stego attacks (for general DNNs) by up to 98.1%, while achieving the same attack success rate (ASR) of 100%. More importantly, SASER improves ASR on quantized models from 0 to 100% in all settings. We appeal for investigations on countermeasures against SASER in view of the significant attack effectiveness.

Medical order extraction is essential for structuring actionable clinical information, supporting decision-making, and enabling downstream applications such as documentation and workflow automation. Orders may be embedded in diverse sources, including electronic health records, discharge summaries, and multi-turn doctor-patient dialogues, and can span categories such as medications, laboratory tests, imaging studies, and follow-up actions. The MEDIQA-OE 2025 shared task focuses on extracting structured medical orders from extended conversational transcripts, requiring the identification of order type, description, reason, and provenance. We present the MasonNLP submission, which ranked 5th among 17 participating teams with 105 total submissions. Our approach uses a general-purpose, instruction-tuned LLaMA-4 17B model without domain-specific fine-tuning, guided by a single in-context example. This few-shot configuration achieved an average F1 score of 37.76, with notable improvements in reason and provenance accuracy. These results demonstrate that large, non-domain-specific LLMs, when paired with effective prompt engineering, can serve as strong, scalable baselines for specialized clinical NLP tasks.

The deployment of large language models (LLMs) is increasingly constrained by memory and latency bottlenecks, motivating the need for quantization techniques that flexibly balance accuracy and efficiency. Recent work has introduced multi-precision models, which enable inference at multiple precisions within a single model depending on runtime constraints. To support such flexibility, quantized weights are often stored as bit-planes, where hardware efficiency improves when the compute operates directly at the bit-plane level and activates only the precision required by each request. In this work, we present AnyBCQ, a hardware-friendly multi-precision extension of Binary-Coded Quantization (BCQ) that supports direct bit-plane operations. By representing weights as binary bit-planes with corresponding scale factors, AnyBCQ enables bit-plane-level computation and maps naturally to accelerator-friendly, bit-parallel arithmetic. Our progressive precision expansion mechanism incrementally refines scaling factors while reusing previously assigned binary codes, yielding monotonic improvements in accuracy as additional bits are enabled. We further co-design a specialized kernel that exploits the BCQ structure to support dynamic per-request precision selection with negligible overhead. Experiments on recent LLMs demonstrate that AnyBCQ significantly narrows the accuracy drop in the low-bit regime (e.g. 2-bit), remains competitive at higher precision, and achieves throughput gains of up to 3.0x over half precision and 1.2x over state-of-the-art multi-precision methods. By aligning algorithmic flexibility with hardware efficiency, AnyBCQ provides a practical foundation for multi-precision LLM deployment across diverse service-level objectives.

Modern Internet of Things (IoT) systems generate massive, heterogeneous multivariate time series data. Accurate Multivariate Time Series Forecasting (MTSF) of such data is critical for numerous applications. However, existing methods almost universally employ a shared embedding layer that processes all channels identically, creating a representational bottleneck that obscures valuable channel-specific information. To address this challenge, we introduce a Shared-Auxiliary Embedding (SAE) framework that decomposes the embedding into a shared base component capturing common patterns and channel-specific auxiliary components modeling unique deviations. Within this decomposition, we \rev{empirically observe} that the auxiliary components tend to exhibit low-rank and clustering characteristics, a structural pattern that is significantly less apparent when using purely independent embeddings. Consequently, we design LightSAE, a parameter-efficient embedding module that operationalizes these observed characteristics through low-rank factorization and a shared, gated component pool. Extensive experiments across 9 IoT-related datasets and 4 backbone architectures demonstrate LightSAE's effectiveness, achieving MSE improvements of up to 22.8\% with only 4.0\% parameter increase.

Medical image segmentation annotation suffers from inter-rater variability (IRV) due to differences in annotators' expertise and the inherent blurriness of medical images. Standard approaches that simply average expert labels are flawed, as they discard the valuable clinical uncertainty revealed in disagreements. We introduce a fundamentally new approach with our group decision simulation framework, which works by mimicking the collaborative decision-making process of a clinical panel. Under this framework, an Expert Signature Generator (ESG) learns to represent individual annotator styles in a unique latent space. A Simulated Consultation Module (SCM) then intelligently generates the final segmentation by sampling from this space. This method achieved state-of-the-art results on challenging CBCT and MRI datasets (92.11% and 90.72% Dice scores). By treating expert disagreement as a useful signal instead of noise, our work provides a clear path toward more robust and trustworthy AI systems for healthcare.

Digital communication has become the cornerstone of modern interaction, enabling rapid, accessible, and interactive exchanges. However, individuals with lower academic literacy often face significant barriers, exacerbating the "digital divide". In this work, we introduce a novel, universal ideographic metalanguage designed as an innovative communication framework that transcends academic, linguistic, and cultural boundaries. Our approach leverages principles of Neuro-symbolic AI, combining neural-based large language models (LLMs) enriched with world knowledge and symbolic knowledge heuristics grounded in the linguistic theory of Natural Semantic Metalanguage (NSM). This enables the semantic decomposition of complex ideas into simpler, atomic concepts. Adopting a human-centric, collaborative methodology, we engaged over 200 semi-literate participants in defining the problem, selecting ideographs, and validating the system. With over 80\% semantic comprehensibility, an accessible learning curve, and universal adaptability, our system effectively serves underprivileged populations with limited formal education.

As the demand for comprehensive evaluations of diverse model capabilities steadily increases, benchmark suites have correspondingly grown significantly in scale. Despite notable advances in redundancy reduction and subset-level performance prediction, a systematic framework that effectively integrates these methods to ensure both prediction accuracy and ranking consistency is still largely elusive. In this paper, we first perform a sample-level analysis of benchmark redundancy and identify several highly similar samples that can be eliminated. Besides, we frame benchmark compression as an optimization problem with the aim of score reconstruction. Building on these, we then propose EssenceBench, a coarse-to-fine framework utilizing an iterative Genetic Algorithm (GA), which takes the advantages of fitness-based subset search and attribution-based sample search. Compared to previous methods, our approach yields superior compression results with lower reconstruction error and markedly higher efficiency. In particular, on the HellaSwag benchmark (10K samples), our method preserves the ranking of all models shifting within 5% using 25x fewer samples, and achieves 95% ranking preservation shifting within 5% using only 200x fewer samples.

Large language models (LLMs) offer a generalizable approach for modeling patient trajectories, but suffer from the long and noisy nature of electronic health records (EHR) data in temporal reasoning. To address these challenges, we introduce Traj-CoA, a multi-agent system involving chain-of-agents for patient trajectory modeling. Traj-CoA employs a chain of worker agents to process EHR data in manageable chunks sequentially, distilling critical events into a shared long-term memory module, EHRMem, to reduce noise and preserve a comprehensive timeline. A final manager agent synthesizes the worker agents' summary and the extracted timeline in EHRMem to make predictions. In a zero-shot one-year lung cancer risk prediction task based on five-year EHR data, Traj-CoA outperforms baselines of four categories. Analysis reveals that Traj-CoA exhibits clinically aligned temporal reasoning, establishing it as a promisingly robust and generalizable approach for modeling complex patient trajectories.

Safety alignment in large language models (LLMs) induces over-refusals -- where LLMs decline benign requests due to aggressive safety filters. We analyze this phenomenon in retrieval-augmented generation (RAG), where both the query intent and retrieved context properties influence refusal behavior. We construct RagRefuse, a domain-stratified benchmark spanning medical, chemical, and open domains, pairing benign and harmful queries with controlled context contamination patterns and sizes. Our analysis shows that context arrangement / contamination, domain of query and context, and harmful-text density trigger refusals even on benign queries, with effects depending on model-specific alignment choices. To mitigate over-refusals, we introduce \textsc{SafeRAG-Steering}, a model-centric embedding intervention that steers the embedding regions towards the confirmed safe, non-refusing output regions at inference time. This reduces over-refusals in contaminated RAG pipelines while preserving legitimate refusals.

Retrieval-augmented generation (RAG) systems trained using reinforcement learning (RL) with reasoning are hampered by inefficient context management, where long, noisy retrieved documents increase costs and degrade performance. We introduce RECON (REasoning with CONdensation), a framework that integrates an explicit summarization module to compress evidence within the reasoning loop. Our summarizer is trained via a two-stage process: relevance pretraining on QA datasets, followed by multi-aspect distillation from proprietary LLMs to ensure factuality and clarity. Integrated into the Search-R1 pipeline, RECON reduces total context length by 35\%, leading to improved training speed and inference latency, while simultaneously improving RAG performance on downstream QA benchmarks. Notably, it boosts the average EM score of the 3B model by 14.5\% and the 7B model by 3.0\%, showing particular strength in multi-hop QA. RECON demonstrates that learned context compression is essential for building practical, scalable, and performant RAG systems. Our code implementation is made available at https://github.com/allfornancy/RECON.