News

Have you ever felt “velvetmist”?  It’s a “complex and subtle emotion that elicits feelings of comfort, serenity, and a gentle sense of floating.” It’s peaceful, but more ephemeral and intangible than contentment. It might be evoked by the sight of a sunset or a moody, low-key album.   If you haven’t ever felt this sensation—or even heard of it—that’s not surprising. A Reddit user named noahjeadie generated it with ChatGPT, along with advice on how to evoke the feeling. With the right essential oils and soundtrack, apparently, you too can feel like “a soft fuzzy draping ghost floating through a lavender suburb.” Don’t scoff: Researchers say more and more terms for these “neo-­emotions” are showing up online, describing new dimensions and aspects of feeling. Velvetmist was a key example in a journal article about the phenomenon published in July 2025. But most neo-emotions aren’t the inventions of emo artificial intelligences. Humans come up with them, and they’re part of a big change in the way researchers are thinking about feelings, one that emphasizes how people continuously spin out new ones in response to a changing world.  Velvetmist might’ve been a chatbot one-off, but it’s not unique. The sociologist Marci Cottingham—whose 2024 paper got this vein of neo-emotion research started—cites many more new terms in circulation. There’s “Black joy” (Black people celebrating embodied pleasure as a form of political resistance), “trans euphoria” (the joy of having one’s gender identity affirmed and celebrated), “eco-anxiety” (the hovering fear of climate disaster), “hypernormalization” (the surreal pressure to continue performing mundane life and labor under capitalism during a global pandemic or fascist takeover), and the sense of “doom” found in “doomer” (one who is relentlessly pessimistic) or “doomscrolling” (being glued to an endless feed of bad news in an immobilized state combining apathy and dread).  Of course, emotional vocabulary is always evolving. During the Civil War, doctors used the centuries-old term “nostalgia,” combining the Greek words for “returning home”and “pain,” to describe a sometimes fatal set of symptoms suffered by soldiers—a condition we’d probably describe today as post-traumatic stress disorder. Now nostalgia’s meaning has mellowed and faded to a gentle affection for an old cultural product or vanished way of life. And people constantly import emotion words from other cultures when they’re convenient or evocative—like hygge (the Danish word for friendly coziness) or kvell (a Yiddish term for brimming over with happy pride).  Cottingham believes that neo-­emotions are proliferating as people spend more of their lives online. These coinages help us relate to one another and make sense of our experiences, and they get a lot of engagement on social media. So even when a neo-emotion is just a subtle variation on, or combination of, existing feelings, getting super-specific about those feelings helps us reflect and connect with other people. “These are potentially signals that tell us about our place in the world,” she says.  These neo-emotions are part of a paradigm shift in emotion science. For decades, researchers argued that humans all share a set of a half-dozen or so basic emotions. But over the last decade, Lisa Feldman Barrett, a clinical psychologist at Northeastern University, has become one of the most cited scientists in the world for work demonstrating otherwise. By using tools like advanced brain imaging and studying babies and people from relatively isolated cultures, she has concluded there’s no such thing as a basic emotional palette. The way we experience and talk about our feelings is culturally determined. “How do you know what anger and sadness and fear are? Because somebody taught you,” Barrett says.  If there are no true “basic” biological emotions, this puts more emphasis on social and cultural variations in how we interpret our experiences. And these interpretations can change over time. “As a sociologist, we think of all emotions as created,” Cottingham says. Just like any other tool humans make and use, “emotions are a practical resource people are using as they navigate the world.”  Some neo-emotions, like velvetmist, might be mere novelties. Barrett playfully suggests “chiplessness” to describe the combined hunger, frustration, and relief of getting to the bottom of the bag. But others, like eco-anxiety and Black joy, can take on a life of their own and help galvanize social movements.   Both reading about and crafting your own neo-emotions, with or without chatbot assistance, could be surprisingly helpful. Lots of research supports the benefits of emotional granularity. Basically, the more detailed and specific words you can use to describe your emotions, both positive and negative, the better.  Researchers analogize this “emodiversity” to biodiversity or cultural diversity, arguing that a more diverse world is more enriched. It turns out that people who exhibit higher emotional granularity go to the doctor less frequently, spend fewer days hospitalized for illness, and are less likely to drink when stressed, drive recklessly, or smoke cigarettes. And many studies show emodiversity is a skill that, with training, people can develop at any age. Just imagine cruising into this sweet, comforting future. Is the idea giving you a certain dreamy thrill? Are you sure you’ve never felt velvetmist? Anya Kamenetz is a freelance education reporter who writes the Substack newsletter The Golden Hour.

arXiv:2512.23280v1 Announce Type: new Abstract: E-commerce live streaming in China, particularly on platforms like Douyin, has become a major sales channel, but hosts often use morphs to evade scrutiny and engage in false advertising. This study introduces the Live Auditory Morph Resolution (LiveAMR) task to detect such violations. Unlike previous morph research focused on text-based evasion in social media and underground industries, LiveAMR targets pronunciation-based evasion in health and medical live streams. We constructed the first LiveAMR dataset with 86,790 samples and developed a method to transform the task into a text-to-text generation problem. By leveraging large language models (LLMs) to generate additional training data, we improved performance and demonstrated that morph resolution significantly enhances live streaming regulation.

LLMRouter is an open source routing library from the U Lab at the University of Illinois Urbana Champaign that treats model selection as a first class system problem. It sits between applications and a pool of LLMs and chooses a model for each query based on task complexity, quality targets, and cost, all exposed through a unified Python API and CLI. The project ships with more than 16 routing models, a data generation pipeline over 11 benchmarks, and a plugin system for custom routers. Router families and supported models LLMRouter organizes routing algorithms into four families, Single-Round Routers, Multi-Round Routers, Personalized Routers, and Agentic Routers. Single round routers include knnrouter, svmrouter, mlprouter, mfrouter, elorouter, routerdc, automix, hybrid_llm, graphrouter, causallm_router, and the baselines smallest_llm and largest_llm. These models implement strategies such as k nearest neighbors, support vector machines, multilayer perceptrons, matrix factorization, Elo rating, dual contrastive learning, automatic model mixing, and graph based routing. Multi round routing is exposed through router_r1, a pre trained instance of Router R1 integrated into LLMRouter. Router R1 formulates multi LLM routing and aggregation as a sequential decision process where the router itself is an LLM that alternates between internal reasoning steps and external model calls. It is trained with reinforcement learning using a rule based reward that balances format, outcome, and cost. In LLMRouter, router_r1 is available as an extra installation target with pinned dependencies tested on vllm==0.6.3 and torch==2.4.0. Personalized routing is handled by gmtrouter, described as a graph based personalized router with user preference learning. GMTRouter represents multi turn user LLM interactions as a heterogeneous graph over users, queries, responses, and models. It runs a message passing architecture over this graph to infer user specific routing preferences from few shot interaction data, and experiments show accuracy and AUC gains over non personalized baselines. Agentic routers in LLMRouter extend routing to multi step reasoning workflows. knnmultiroundrouter uses k nearest neighbor reasoning over multi turn traces and is intended for complex tasks. llmmultiroundrouter exposes an LLM based agentic router that performs multi step routing without its own training loop. These agentic routers share the same configuration and data formats as the other router families and can be swapped through a single CLI flag. Data generation pipeline for routing datasets LLMRouter ships with a full data generation pipeline that turns standard benchmarks and LLM outputs into routing datasets. The pipeline supports 11 benchmarks, Natural QA, Trivia QA, MMLU, GPQA, MBPP, HumanEval, GSM8K, CommonsenseQA, MATH, OpenBookQA, and ARC Challenge. It runs in three explicit stages. First, data_generation.py extracts queries and ground truth labels and creates train and test JSONL splits. Second, generate_llm_embeddings.py builds embeddings for candidate LLMs from metadata. Third, api_calling_evaluation.py calls LLM APIs, evaluates responses, and fuses scores with embeddings into routing records. (GitHub) The pipeline outputs query files, LLM embedding JSON, query embedding tensors, and routing data JSONL files. A routing entry includes fields such as task_name, query, ground_truth, metric, model_name, response, performance, embedding_id, and token_num. Configuration is handled entirely through YAML, so engineers point the scripts to new datasets and candidate model lists without modifying code. Chat interface and plugin system For interactive use, llmrouter chat launches a Gradio based chat frontend over any router and configuration. The server can bind to a custom host and port and can expose a public sharing link. Query modes control how routing sees context. current_only uses only the latest user message, full_context concatenates the dialogue history, and retrieval augments the query with the top k similar historical queries. The UI visualizes model choices in real time and is driven by the same router configuration used for batch inference. LLMRouter also provides a plugin system for custom routers. New routers live under custom_routers, subclass MetaRouter, and implement route_single and route_batch. Configuration files under that directory define data paths, hyperparameters, and optional default API endpoints. Plugin discovery scans the project custom_routers folder, a ~/.llmrouter/plugins directory, and any extra paths in the LLMROUTER_PLUGINS environment variable. Example custom routers include randomrouter, which selects a model at random, and thresholdrouter, which is a trainable router that estimates query difficulty. Key Takeaways Routing as a first class abstraction: LLMRouter is an open source routing layer from UIUC that sits between applications and heterogeneous LLM pools and centralizes model selection as a cost and quality aware prediction task rather than ad hoc scripts. Four router families covering 16 plus algorithms: The library standardizes more than 16 routers into four families, single round, multi round, personalized, and agentic, including knnrouter, graphrouter, routerdc, router_r1, and gmtrouter, all exposed through a unified config and CLI. Multi round RL routing via Router R1: router_r1 integrates the Router R1 framework, where an LLM router interleaves internal “think” steps with external “route” calls and is trained with a rule based reward that combines format, outcome, and cost to optimize performance cost trade offs. Graph based personalization with GMTRouter: gmtrouter models users, queries, responses and LLMs as nodes in a heterogeneous graph and uses message passing to learn user specific routing preferences from few shot histories, achieving up to around 21% accuracy gains and substantial AUC improvements over strong baselines. End to end pipeline and extensibility: LLMRouter provides a benchmark driven data pipeline, CLI for training and inference, a Gradio chat UI, centralized API key handling, and a plugin system based on MetaRouter that allows teams to register custom routers while reusing the same routing datasets and infrastructure. Check out the GitHub Repo and Technical details. Also, feel free to follow us on Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. The post Meet LLMRouter: An Intelligent Routing System designed to Optimize LLM Inference by Dynamically Selecting the most Suitable Model for Each Query appeared first on MarkTechPost.

arXiv:2512.22955v1 Announce Type: new Abstract: Recent advancements have shown that reinforcement learning (RL) can substantially improve the reasoning abilities of large language models (LLMs). The effectiveness of such RL training, however, depends critically on the exploration space defined by the pre-trained model’s token-output distribution. In this paper, we revisit the standard cross-entropy loss, interpreting it as a specific instance of policy gradient optimization applied within a single-step episode. To systematically study how the pre-trained distribution shapes the exploration potential for subsequent RL, we propose a generalized pre-training objective that adapts on-policy RL principles to supervised learning. By framing next-token prediction as a stochastic decision process, we introduce a reward-shaping strategy that explicitly balances diversity and precision. Our method employs a positive reward scaling factor to control probability concentration on ground-truth tokens and a rank-aware mechanism that treats high-ranking and low-ranking negative tokens asymmetrically. This allows us to reshape the pre-trained token-output distribution and investigate how to provide a more favorable exploration space for RL, ultimately enhancing end-to-end reasoning performance. Contrary to the intuition that higher distribution entropy facilitates effective exploration, we find that imposing a precision-oriented prior yields a superior exploration space for RL.

arXiv:2512.22455v1 Announce Type: cross Abstract: Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method. However, its linear adaptation process limits its expressive power. This means there is a gap between the expressive power of linear training and non-linear training. To bridge this gap, we propose AFA-LoRA, a novel training strategy that brings non-linear expressivity to LoRA while maintaining its seamless mergeability. Our key innovation is an annealed activation function that transitions from a non-linear to a linear transformation during training, allowing the adapter to initially adopt stronger representational capabilities before converging to a mergeable linear form. We implement our method on supervised fine-tuning, reinforcement learning, and speculative decoding. The results show that AFA-LoRA reduces the performance gap between LoRA and full-parameter training. This work enables a more powerful and practical paradigm of parameter-efficient adaptation.

arXiv:2512.22966v1 Announce Type: new Abstract: Large Language Models (LLMs) have demonstrated promise in medical knowledge assessments, yet their practical utility in real-world clinical decision-making remains underexplored. In this study, we evaluated the performance of three state-of-the-art LLMs-ChatGPT-4o, Gemini 1.5 Pro, and LIama 3.3 70B-in clinical decision support across the entire clinical reasoning workflow of a typical patient encounter. Using 36 case studies, we first assessed LLM’s out-of-the-box performance across five key sequential clinical decision-making tasks under two temperature settings (default vs. zero): differential diagnosis, essential immediate steps, relevant diagnostic testing, final diagnosis, and treatment recommendation. All models showed high variability by task, achieving near-perfect accuracy in final diagnosis, poor performance in relevant diagnostic testing, and moderate performance in remaining tasks. Furthermore, ChatGPT performed better under the zero temperature, whereas LIama showed stronger performance under the default temperature. Next, we assessed whether prompt engineering could enhance LLM performance by applying variations of the MedPrompt framework, incorporating targeted and random dynamic few-shot learning. The results demonstrate that prompt engineering is not a one-size-fit-all solution. While it significantly improved the performance on the task with lowest baseline accuracy (relevant diagnostic testing), it was counterproductive for others. Another key finding was that the targeted dynamic few-shot prompting did not consistently outperform random selection, indicating that the presumed benefits of closely matched examples may be counterbalanced by loss of broader contextual diversity. These findings suggest that the impact of prompt engineering is highly model and task-dependent, highlighting the need for tailored, context-aware strategies for integrating LLMs into healthcare.

arXiv:2508.07185v2 Announce Type: replace Abstract: Large Language Models (LLMs) suffer from a critical limitation: their knowledge is static and quickly becomes outdated. Retraining these massive models is computationally prohibitive, while existing knowledge editing techniques can be slow and may introduce unforeseen side effects. To address this, we propose DySK-Attn, a novel framework that enables LLMs to efficiently integrate real-time knowledge from a dynamic external source. Our approach synergizes an LLM with a dynamic Knowledge Graph (KG) that can be updated instantaneously. The core of our framework is a sparse knowledge attention mechanism, which allows the LLM to perform a coarse-to-fine grained search, efficiently identifying and focusing on a small, highly relevant subset of facts from the vast KG. This mechanism avoids the high computational cost of dense attention over the entire knowledge base and mitigates noise from irrelevant information. We demonstrate through extensive experiments on time-sensitive question-answering tasks that DySK-Attn significantly outperforms strong baselines, including standard Retrieval-Augmented Generation (RAG) and model editing techniques, in both factual accuracy for updated knowledge and computational efficiency. Our framework offers a scalable and effective solution for building LLMs that can stay current with the ever-changing world.

arXiv:2508.10426v2 Announce Type: replace Abstract: Large language models (LLMs) are limited by substantial computational cost. We introduce a “computational economics” framework that treats an LLM as an internal economy of resource-constrained agents (attention heads and neuron blocks) that must allocate scarce computation to maximize task utility. First, we show empirically that when computation is scarce, standard LLMs reallocate attention toward high-value tokens while preserving accuracy. Building on this observation, we propose an incentive-driven training paradigm that augments the task loss with a differentiable computation cost term, encouraging sparse and efficient activations. On GLUE (MNLI, STS-B, CoLA) and WikiText-103, the method yields a family of models that trace a Pareto frontier and consistently dominate post-hoc pruning; for a similar accuracy we obtain roughly a forty percent reduction in FLOPS and lower latency, together with more interpretable attention patterns. These results indicate that economic principles offer a principled route to designing efficient, adaptive, and more transparent LLMs under strict resource constraints.

arXiv:2411.02455v2 Announce Type: replace Abstract: Course evaluation plays a critical role in ensuring instructional quality and guiding curriculum development in higher education. However, traditional evaluation methods, such as student surveys, classroom observations, and expert reviews, are often constrained by subjectivity, high labor costs, and limited scalability. With recent advancements in large language models (LLMs), new opportunities have emerged for generating consistent, fine-grained, and scalable course evaluations. This study investigates the use of three representative LLMs for automated course evaluation at both the micro level (classroom discussion analysis) and the macro level (holistic course review). Using classroom interaction transcripts and a dataset of 100 courses from a major institution in China, we demonstrate that LLMs can extract key pedagogical features and generate structured evaluation results aligned with expert judgement. A fine-tuned version of Llama shows superior reliability, producing score distributions with greater differentiation and stronger correlation with human evaluators than its counterparts. The results highlight three major findings: (1) LLMs can reliably perform systematic and interpretable course evaluations at both the micro and macro levels; (2) fine-tuning and prompt engineering significantly enhance evaluation accuracy and consistency; and (3) LLM-generated feedback provides actionable insights for teaching improvement. These findings illustrate the promise of LLM-based evaluation as a practical tool for supporting quality assurance and educational decision-making in large-scale higher education settings.

arXiv:2509.10436v2 Announce Type: replace Abstract: To optimize the reasoning and problem-solving capabilities of Large Language Models (LLMs), we propose a novel cloud-edge collaborative architecture that enables a structured multi-agent prompting framework. This framework comprises three specialized components: GuideLLM, a lightweight model deployed at the edge to provide methodological guidance; SolverLLM, a more powerful model hosted in the cloud and responsible for generating code solutions; and JudgeLLM, an automated evaluator for assessing solution correctness and quality. To evaluate and demonstrate the effectiveness of this architecture in realistic settings, we introduce RefactorCoderQA, a comprehensive benchmark designed to evaluate and enhance the performance of LLMs across multi-domain coding tasks. Motivated by the limitations of existing benchmarks, RefactorCoderQA systematically covers multiple technical domains, including Software Engineering, Data Science, Machine Learning, and Natural Language Processing, using authentic coding challenges sourced from Stack Overflow. We propose RefactorCoder-MoE, a fine-tuned mixture-of-experts (MoE) code language model based on DeepSeek-Coder-7B-Instruct, adapted to the RefactorCoderQA benchmark using QLoRA for domain-specific coding question answering. Extensive experiments demonstrate that RefactorCoder-MoE achieves strong and competitive performance, significantly outperforming all evaluated open-source and commercial baselines, with an overall accuracy of 76.84%.