Rapid developments of large language models have revolutionized many NLP tasks for English data. Unfortunately, the models and their evaluations for low-resource languages are being overlooked, especially for languages in South Asia. Although there are more than 650 languages in South Asia, many of them either have very limited computational resources or are missing from existing language models. Thus, a concrete question to be answered is: Can we assess the current stage and challenges to inform our NLP community and facilitate model developments for South Asian languages? In this survey, we have comprehensively examined current efforts and challenges of NLP models for South Asian languages by retrieving studies since 2020, with a focus on transformer-based models, such as BERT, T5, & GPT. We present advances and gaps across 3 essential aspects: data, models, & tasks, such as available data sources, fine-tuning strategies, & domain applications. Our findings highlight substantial issues, including missing data in critical domains (e.g., health), code-mixing, and lack of standardized evaluation benchmarks. Our survey aims to raise awareness within the NLP community for more targeted data curation, unify benchmarks tailored to cultural and linguistic nuances of South Asia, and encourage an equitable representation of South Asian languages. The complete list of resources is available at: https://github.com/trust-nlp/LM4SouthAsia-Survey.

BACKGROUND: Medical large language models (LLMS) have demonstrated remarkable performance in answering medical examinations. However, the extent to which this high performance is transferable to medical questions in Spanish and from a Latin American country remains unexplored. This knowledge is crucial as LLM-based medical applications gain traction in Latin America. AIMS: to build a dataset of questions from medical examinations taken by Peruvian physicians pursuing specialty training; to fine-tune a LLM on this dataset; to evaluate and compare the performance in terms of accuracy between vanilla LLMs and the fine-tuned LLM. METHODS: We curated PeruMedQA, a multiple-choice question-answering (MCQA) datasets containing 8,380 questions spanning 12 medical domains (2018-2025). We selected eight medical LLMs including medgemma-4b-it and medgemma-27b-text-it, and developed zero-shot task-specific prompts to answer the questions appropriately. We employed parameter-efficient fine tuning (PEFT)and low-rant adaptation (LoRA) to fine-tune medgemma-4b-it utilizing all questions except those from 2025 (test set). RESULTS: medgemma-27b-text-it outperformed all other models, achieving a proportion of correct answers exceeding 90% in several instances. LLMs with <10 billion parameters exhibited <60% of correct answers, while some exams yielded results <50%. The fine-tuned version of medgemma-4b-it emerged victorious agains all LLMs with <10 billion parameters and rivaled a LLM with 70 billion parameters across various examinations. CONCLUSIONS: For medical AI application and research that require knowledge bases from Spanish-speaking countries and those exhibiting similar epidemiological profiles to Peru's, interested parties should utilize medgemma-27b-text-it or a fine-tuned version of medgemma-4b-it.

Persian names present unique challenges for natural language processing applications, particularly in gender detection and digital identity creation, due to transliteration inconsistencies and cultural-specific naming patterns. Existing tools exhibit significant performance degradation on Persian names, while the scarcity of comprehensive datasets further compounds these limitations. To address these challenges, the present research introduces PNGT-26K, a comprehensive dataset of Persian names, their commonly associated gender, and their English transliteration, consisting of approximately 26,000 tuples. As a demonstration of how this resource can be utilized, we also introduce two frameworks, namely Open Gender Detection and Nominalist. Open Gender Detection is a production-grade, ready-to-use framework for using existing data from a user, such as profile photo and name, to give a probabilistic guess about the person's gender. Nominalist, the second framework introduced by this paper, utilizes agentic AI to help users choose a username for their social media accounts on any platform. It can be easily integrated into any website to provide a better user experience. The PNGT-26K dataset, Nominalist and Open Gender Detection frameworks are publicly available on Github.

New technologies in generative AI can enable deeper analysis into our nation's supply chains but truly informative insights require the continual updating and aggregation of massive data in a timely manner. Large Language Models (LLMs) offer unprecedented analytical opportunities however, their knowledge base is constrained to the models' last training date, rendering these capabilities unusable for organizations whose mission impacts rely on emerging and timely information. This research proposes an innovative approach to supply chain analysis by integrating emerging Retrieval-Augmented Generation (RAG) preprocessing and retrieval techniques with advanced web-scraping technologies. Our method aims to reduce latency in incorporating new information into an augmented-LLM, enabling timely analysis of supply chain disruptors. Through experimentation, this study evaluates the combinatorial effects of these techniques towards timeliness and quality trade-offs. Our results suggest that in applying RAG systems to supply chain analysis, fine-tuning the embedding retrieval model consistently provides the most significant performance gains, underscoring the critical importance of retrieval quality. Adaptive iterative retrieval, which dynamically adjusts retrieval depth based on context, further enhances performance, especially on complex supply chain queries. Conversely, fine-tuning the LLM yields limited improvements and higher resource costs, while techniques such as downward query abstraction significantly outperforms upward abstraction in practice.

Marmoset monkeys exhibit complex vocal communication, challenging the view that nonhuman primates vocal communication is entirely innate, and show similar features of human speech, such as vocal labeling of others and turn-taking. Studying their vocal communication offers a unique opportunity to link it with brain activity-especially given the difficulty of accessing the human brain in speech and language research. Since Marmosets communicate primarily through vocalizations, applying standard LLM approaches is not straightforward. We introduce Generative Marmoset Spoken Language Modeling (GmSLM), an optimized spoken language model pipeline for Marmoset vocal communication. We designed a novel zero-shot evaluation metrics using unsupervised in-the-wild data, alongside weakly labeled conversational data, to assess GmSLM and demonstrate its advantage over a basic human-speech-based baseline. GmSLM generated vocalizations closely matched real resynthesized samples acoustically and performed well on downstream tasks. Despite being fully unsupervised, GmSLM effectively distinguish real from artificial conversations and may support further investigations of the neural basis of vocal communication and provides a practical framework linking vocalization and brain activity. We believe GmSLM stands to benefit future work in neuroscience, bioacoustics, and evolutionary biology. Samples are provided under: pages.cs.huji.ac.il/adiyoss-lab/GmSLM.

Targeting the issues of "shortcuts" and insufficient contextual understanding in complex cross-modal reasoning of multimodal large models, this paper proposes a zero-shot multimodal reasoning component guided by human-like cognitive strategies centered on an "intent sketch". The component comprises a plug-and-play three-module pipeline-Intent Perceiver, Strategy Generator, and Strategy Selector-that explicitly constructs a "understand-plan-select" cognitive process. By generating and filtering "intent sketch" strategies to guide the final reasoning, it requires no parameter fine-tuning and achieves cross-model transfer solely through in-context engineering. Information-theoretic analysis shows that this process can reduce conditional entropy and improve information utilization efficiency, thereby suppressing unintended shortcut reasoning. Experiments on IntentBench, WorldSense, and Daily-Omni validate the method's generality and robust gains; compared with their respective baselines, the complete "three-module" scheme yields consistent improvements across different reasoning engines and pipeline combinations, with gains up to approximately 9.51 percentage points, demonstrating the practical value and portability of the "intent sketch" reasoning component in zero-shot scenarios.

Every day, our inboxes are flooded with unsolicited emails, ranging between annoying spam to more subtle phishing scams. Unfortunately, despite abundant prior efforts proposing solutions achieving near-perfect accuracy, the reality is that countering malicious emails still remains an unsolved dilemma. This "open problem" paper carries out a critical assessment of scientific works in the context of phishing email detection. First, we focus on the benchmark datasets that have been used to assess the methods proposed in research. We find that most prior work relied on datasets containing emails that -- we argue -- are not representative of current trends, and mostly encompass the English language. Based on this finding, we then re-implement and re-assess a variety of detection methods reliant on machine learning (ML), including large-language models (LLM), and release all of our codebase -- an (unfortunately) uncommon practice in related research. We show that most such methods achieve near-perfect performance when trained and tested on the same dataset -- a result which intrinsically hinders development (how can future research outperform methods that are already near perfect?). To foster the creation of "more challenging benchmarks" that reflect current phishing trends, we propose E-PhishGEN, an LLM-based (and privacy-savvy) framework to generate novel phishing-email datasets. We use our E-PhishGEN to create E-PhishLLM, a novel phishing-email detection dataset containing 16616 emails in three languages. We use E-PhishLLM to test the detectors we considered, showing a much lower performance than that achieved on existing benchmarks -- indicating a larger room for improvement. We also validate the quality of E-PhishLLM with a user study (n=30). To sum up, we show that phishing email detection is still an open problem -- and provide the means to tackle such a problem by future research.

The rapid advancement of speech-to-speech (S2S) large language models (LLMs) has significantly improved real-time spoken interaction. However, current evaluation frameworks remain inadequate for assessing performance in complex, multi-turn dialogues. To address this, we introduce MTalk-Bench, a multi-turn S2S benchmark covering three core dimensions: Semantic Information, Paralinguistic Information, and Ambient Sound. Each dimension includes nine realistic scenarios, along with targeted tasks to assess specific capabilities such as reasoning. Our dual-method evaluation framework combines Arena-style evaluation (pairwise comparison) and Rubrics-based evaluation (absolute scoring) for relative and absolute assessment. The benchmark includes both model and human outputs, evaluated by human evaluators and LLMs. Experimental results reveal two sets of findings. Overall performance of S2S LLMs: (1) models excel at semantic information processing yet underperform on paralinguistic information and ambient sounds perception; (2) models typically regain coherence by increasing response length, sacrificing efficiency in multi-turn dialogues; (3) modality-aware, task-specific designs outperform brute scaling. Evaluation framework and reliability: (1) Arena and Rubrics yield consistent, complementary rankings, but reliable distinctions emerge only when performance gaps are large; (2) LLM-as-a-judge aligns with humans when gaps are clear or criteria explicit, but exhibits position and length biases and is reliable on nonverbal evaluation only with text annotations. These results highlight current limitations in S2S evaluation and the need for more robust, speech-aware assessment frameworks.

Abstract--Large Language Model (LLM) inference is increasingly constrained by memory bandwidth, with frequent access to the key-value (KV) cache dominating data movement. While attention sparsity reduces some memory traffic, the relevance of past tokens varies over time, requiring the full KV cache to remain accessible and sustaining pressure on both bandwidth and capacity. This work investigates dynamic KV cache placement across such systems to maximize aggregated bandwidth utilization under capacity constraints. Rather than proposing a specific scheduling policy, we formulate the placement problem mathematically and derive a theoretical upper bound, revealing substantial headroom for runtime optimization. T o our knowledge, this is the first formal treatment of dynamic KV cache scheduling in heterogeneous memory systems for LLM inference.

Vision-Language Models (VLMs) struggle with complex image annotation tasks, such as emotion classification and context-driven object detection, which demand sophisticated reasoning. Standard Supervised Fine-Tuning (SFT) focuses solely on annotation outcomes, ignoring underlying rationales, while Visual Reinforcement Fine-Tuning (Visual-RFT) produces inconsistent Chains of Thought (CoTs) due to the absence of high-quality, verified CoTs during pre-training. We introduce RISE (Reason-Inspire-Strengthen-Expertise), a two-stage framework to overcome these limitations. In the Reason stage (RISE-CoT), a reinforcement learning-driven "annotation-reasoning-annotation" closed-loop generates visually grounded, logically consistent CoTs by verifying their ability to reconstruct original annotations without direct leakage. The Inspire and Strengthen stage (RISE-R1) leverages a high-quality CoT subset, filtered by RISE-CoT rewards, for supervised fine-tuning, followed by reinforcement fine-tuning to produce interpretable reasoning and accurate annotations, achieving Expertise in complex visual tasks. Evaluated on complex and simple image annotation tasks, RISE-trained Qwen2-VL-2B outperforms SFT and Visual-RFT, achieving robust performance and enhanced explainability. RISE offers a self-supervised solution for advancing VLM reasoning without requiring manually annotated CoTs.Code and resources are available at: https://github.com/HSH55/RISE.