Uncategorized

arXiv:2512.17843v2 Announce Type: replace Abstract: While academic research typically treats Large Language Models (LLM) as generic text generators, they are distinct commercial products with unique interfaces and capabilities that fundamentally shape user behavior. Current datasets obscure this reality by collecting text-only data through uniform interfaces that fail to capture authentic chatbot usage. To address this limitation, we present ShareChat, a large-scale corpus of 142,808 conversations (660,293 turns) sourced directly from publicly shared URLs on ChatGPT, Perplexity, Grok, Gemini, and Claude. ShareChat distinguishes itself by preserving native platform affordances, such as citations and thinking traces, across a diverse collection covering 101 languages and the period from April 2023 to October 2025. Furthermore, ShareChat offers substantially longer context windows and greater interaction depth than prior datasets. To illustrate the dataset’s breadth, we present three case studies: a completeness analysis of intent satisfaction, a citation study of model grounding, and a temporal analysis of engagement rhythms. This work provides the community with a vital and timely resource for understanding authentic user-LLM chatbot interactions in the wild. The dataset will be publicly available.

arXiv:2601.02609v1 Announce Type: cross Abstract: Large language model fine-tuning is bottlenecked by memory: a 7B parameter model requires 84GB–14GB for weights, 14GB for gradients, and 56GB for FP32 optimizer states–exceeding even A100-40GB capacity. We present Chronicals, an open-source training framework achieving 3.51x speedup over Unsloth through four synergistic optimizations: (1) fused Triton kernels eliminating 75% of memory traffic via RMSNorm (7x), SwiGLU (5x), and QK-RoPE (2.3x) fusion; (2) Cut Cross-Entropy reducing logit memory from 5GB to 135MB through online softmax computation; (3) LoRA+ with theoretically-derived 16x differential learning rates between adapter matrices; and (4) Best-Fit Decreasing sequence packing recovering 60-75% of compute wasted on padding. On Qwen2.5-0.5B with A100-40GB, Chronicals achieves 41,184 tokens/second for full fine-tuning versus Unsloth’s 11,736 tokens/second (3.51x). For LoRA at rank 32, we reach 11,699 tokens/second versus Unsloth MAX’s 2,857 tokens/second (4.10x). Critically, we discovered that Unsloth’s reported 46,000 tokens/second benchmark exhibited zero gradient norms–the model was not training. We provide complete mathematical foundations: online softmax correctness proofs, FlashAttention IO complexity bounds O(N^2 d^2 M^{-1}), LoRA+ learning rate derivations from gradient magnitude analysis, and bin-packing approximation guarantees. All implementations, benchmarks, and proofs are available at https://github.com/Ajwebdevs/Chronicals with pip installation via https://pypi.org/project/chronicals/.

arXiv:2601.01778v1 Announce Type: new Abstract: Despite its widespread use, Bengali lacks a robust automated International Phonetic Alphabet (IPA) transcription system that effectively supports both standard language and regional dialectal texts. Existing approaches struggle to handle regional variations, numerical expressions, and generalize poorly to previously unseen words. To address these limitations, we propose BanglaIPA, a novel IPA generation system that integrates a character-based vocabulary with word-level alignment. The proposed system accurately handles Bengali numerals and demonstrates strong performance across regional dialects. BanglaIPA improves inference efficiency by leveraging a precomputed word-to-IPA mapping dictionary for previously observed words. The system is evaluated on the standard Bengali and six regional variations of the DUAL-IPA dataset. Experimental results show that BanglaIPA outperforms baseline IPA transcription models by 58.4-78.7% and achieves an overall mean word error rate of 11.4%, highlighting its robustness in phonetic transcription generation for the Bengali language.

arXiv:2601.02043v1 Announce Type: cross Abstract: Reasoning has long been understood as a pathway between stages of understanding. Proper reasoning leads to understanding of a given subject. This reasoning was conceptualized as a process of understanding in a particular way, i.e., “symbolic reasoning”. Foundational Models (FM) demonstrate that this is not a necessary condition for many reasoning tasks: they can “reason” by way of imitating the process of “thinking out loud”, testing the produced pathways, and iterating on these pathways on their own. This leads to some form of reasoning that can solve problems on its own or with few-shot learning, but appears fundamentally different from human reasoning due to its lack of grounding and common sense, leading to brittleness of the reasoning process. These insights promise to substantially alter our assessment of reasoning and its necessary conditions, but also inform the approaches to safety and robust defences against this brittleness of FMs. This paper offers and discusses several philosophical interpretations of this phenomenon, argues that the previously apt metaphor of the “stochastic parrot” has lost its relevance and thus should be abandoned, and reflects on different normative elements in the safety- and appropriateness-considerations emerging from these reasoning models and their growing capacity.

arXiv:2601.01825v1 Announce Type: new Abstract: Large Language Models (LLMs) have achieved remarkable success in general benchmarks, yet their competence in commodity supply chains (CSCs) — a domain governed by institutional rule systems and feasibility constraints — remains under-explored. CSC decisions are shaped jointly by process stages (e.g., planning, procurement, delivery), variety-specific rules (e.g., contract specifications and delivery grades), and reasoning depth (from retrieval to multi-step analysis and decision selection). We introduce CSCBench, a 2.3K+ single-choice benchmark for CSC reasoning, instantiated through our PVC 3D Evaluation Framework (Process, Variety, and Cognition). The Process axis aligns tasks with SCOR+Enable; the Variety axis operationalizes commodity-specific rule systems under coupled material-information-financial constraints, grounded in authoritative exchange guidebooks/rulebooks and industry reports; and the Cognition axis follows Bloom’s revised taxonomy. Evaluating representative LLMs under a direct prompting setting, we observe strong performance on the Process and Cognition axes but substantial degradation on the Variety axis, especially on Freight Agreements. CSCBench provides a diagnostic yardstick for measuring and improving LLM capabilities in this high-stakes domain.

arXiv:2601.02023v1 Announce Type: new Abstract: Large language models (LLMs) increasingly support very long input contexts. Yet it remains unclear how reliably they extract and infer information at scale. Performance varies with context length and strongly interacts with how information is distributed in real-world corpora. Motivated by these observations, we study how fact placement, corpus-level fact distributions, and Don’t Make It Up prompts influence model behavior. We introduce an extended needle-in-a-haystack benchmark across four production-scale models: Gemini-2.5-flash, ChatGPT-5-mini, Claude-4.5-haiku, and Deepseek-v3.2-chat. Unlike prior work, we separately evaluate literal extraction, logical inference, and hallucination risk. Our study considers both positional effects and realistic distributions of evidence across long contexts, as well as prompts that explicitly discourage fabrication. We find that longer contexts alone do not guarantee better performance and can be detrimental when relevant evidence is diluted or widely dispersed. Performance varies substantially across models: some show severe degradation under realistic conditions, while others remain more robust at longer context lengths. Anti-hallucination (AH) instructions can make some models overly conservative, sharply reducing accuracy in literal extraction and logical inference. While we do not directly compare retrieval-augmented generation (RAG) and cache-augmented generation (CAG), our results suggest many failures stem from ineffective context utilization. Models often struggle to identify and prioritize relevant information even when it is present. These findings have direct practical implications, as enterprise workflows increasingly involve pasting large volumes of unfiltered documents into LLM prompts. Effective context length and model-specific robustness to long contexts are therefore critical for reliable LLM deployment in research and business.

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Sometimes AI feels like a niche topic to write about, but then the holidays happen, and I hear relatives of all ages talking about cases of chatbot-induced psychosis, blaming rising electricity prices on data centers, and asking whether kids should have unfettered access to AI. It’s everywhere, in other words. And people are alarmed. Inevitably, these conversations take a turn: AI is having all these ripple effects now, but if the technology gets better, what happens next? That’s usually when they look at me, expecting a forecast of either doom or hope.  I probably disappoint, if only because predictions for AI are getting harder and harder to make.  Despite that, MIT Technology Review has, I must say, a pretty excellent track record of making sense of where AI is headed. We’ve just published a sharp list of predictions for what’s next in 2026 (where you can read my thoughts on the legal battles surrounding AI), and the predictions on last year’s list all came to fruition. But every holiday season, it gets harder and harder to work out the impact AI will have. That’s mostly because of three big unanswered questions. For one, we don’t know if large language models will continue getting incrementally smarter in the near future. Since this particular technology is what underpins nearly all the excitement and anxiety in AI right now, powering everything from AI companions to customer service agents, its slowdown would be a pretty huge deal. Such a big deal, in fact, that we devoted a whole slate of stories in December to what a new post-AI-hype era might look like.  Number two, AI is pretty abysmally unpopular among the general public. Here’s just one example: Nearly a year ago, OpenAI’s Sam Altman stood next to President Trump to excitedly announce a $500 billion project to build data centers across the US in order to train larger and larger AI models. The pair either did not guess or did not care that many Americans would staunchly oppose having such data centers built in their communities. A year later, Big Tech is waging an uphill battle to win over public opinion and keep on building. Can it win?  The response from lawmakers to all this frustration is terribly confused. Trump has pleased Big Tech CEOs by moving to make AI regulation a federal rather than a state issue, and tech companies are now hoping to codify this into law. But the crowd that wants to protect kids from chatbots ranges from progressive lawmakers in California to the increasingly Trump-aligned Federal Trade Commission, each with distinct motives and approaches. Will they be able to put aside their differences and rein AI firms in?  If the gloomy holiday dinner table conversation gets this far, someone will say: Hey, isn’t AI being used for objectively good things? Making people healthier, unearthing scientific discoveries, better understanding climate change? Well, sort of. Machine learning, an older form of AI, has long been used in all sorts of scientific research. One branch, called deep learning, forms part of AlphaFold, a Nobel Prize–winning tool for protein prediction that has transformed biology. Image recognition models are getting better at identifying cancerous cells.  But the track record for chatbots built atop newer large language models is more modest. Technologies like ChatGPT are quite good at analyzing large swathes of research to summarize what’s already been discovered. But some high-profile reports that these sorts of AI models had made a genuine discovery, like solving a previously unsolved mathematics problem, were bogus. They can assist doctors with diagnoses, but they can also encourage people to diagnose their own health problems without consulting doctors, sometimes with disastrous results.  This time next year, we’ll probably have better answers to my family’s questions, and we’ll have a bunch of entirely new questions too. In the meantime, be sure to read our full piece forecasting what will happen this year, featuring predictions from the whole AI team.

arXiv:2506.01495v5 Announce Type: replace Abstract: Ensuring that Large Language Models (LLMs) align with mainstream human values and ethical norms is crucial for the safe and sustainable development of AI. Current value evaluation and alignment are constrained by Western cultural bias and incomplete domestic frameworks reliant on non-native rules; furthermore, the lack of scalable, rule-driven scenario generation methods makes evaluations costly and inadequate across diverse cultural contexts. To address these challenges, we propose a hierarchical value framework grounded in core Chinese values, encompassing three main dimensions, 12 core values, and 50 derived values. Based on this framework, we construct a large-scale Chinese Value Rule Corpus (C-VARC) containing over 250,000 value rules enhanced and expanded through human annotation. Experimental results demonstrate that scenarios guided by C-VARC exhibit clearer value boundaries and greater content diversity compared to those produced through direct generation. In the evaluation across six sensitive themes (e.g., surrogacy, suicide), seven mainstream LLMs preferred C-VARC generated options in over 70.5% of cases, while five Chinese human annotators showed an 87.5% alignment with C-VARC, confirming its universality, cultural relevance, and strong alignment with Chinese values. Additionally, we construct 400,000 rule-based moral dilemma scenarios that objectively capture nuanced distinctions in conflicting value prioritization across 17 LLMs. Our work establishes a culturally-adaptive benchmarking framework for comprehensive value evaluation and alignment, representing Chinese characteristics.

arXiv:2601.00510v1 Announce Type: cross Abstract: Search in e-Commerce is powered at the core by a structured representation of the inventory, often formulated as a category taxonomy. An important capability in e-Commerce with hierarchical taxonomies is to select a set of relevant leaf categories that are semantically aligned with a given user query. In this scope, we address a fundamental problem of search query categorization in real-world e-Commerce taxonomies. A correct categorization of a query not only provides a way to zoom into the correct inventory space, but opens the door to multiple intent understanding capabilities for a query. A practical and accurate solution to this problem has many applications in e-commerce, including constraining retrieved items and improving the relevance of the search results. For this task, we explore a novel Chain-of-Thought (CoT) paradigm that combines simple tree-search with LLM semantic scoring. Assessing its classification performance on human-judged query-category pairs, relevance tests, and LLM-based reference methods, we find that the CoT approach performs better than a benchmark that uses embedding-based query category predictions. We show how the CoT approach can detect problems within a hierarchical taxonomy. Finally, we also propose LLM-based approaches for query-categorization of the same spirit, but which scale better at the range of millions of queries.

arXiv:2507.11407v2 Announce Type: replace Abstract: This technical report introduces EXAONE 4.0, which integrates a Non-reasoning mode and a Reasoning mode to achieve both the excellent usability of EXAONE 3.5 and the advanced reasoning abilities of EXAONE Deep. To pave the way for the agentic AI era, EXAONE 4.0 incorporates essential features such as agentic tool use, and its multilingual capabilities are extended to support Spanish in addition to English and Korean. The EXAONE 4.0 model series consists of two sizes: a mid-size 32B model optimized for high performance, and a small-size 1.2B model designed for on-device applications. The EXAONE 4.0 demonstrates superior performance compared to open-weight models in its class and remains competitive even against frontier-class models. The models are publicly available for research purposes and can be easily downloaded via https://huggingface.co/LGAI-EXAONE.