LLM

When large language models (LLMs) reason step-by-step using Chain-of-Thought (CoT) prompting, they think out loud. That verbosity improves accuracy—but it’s also a luxury many applications can’t afford. From real-time voice assistants to robotics, excessive token generation slows everything down. The result is a fundamental bottleneck: performance versus efficiency. The paper SynAdapt: Learning Adaptive Reasoning in Large Language Models via Synthetic Continuous Chain-of-Thought offers a clever solution. Rather than generating verbose natural language steps, SynAdapt trains LLMs to reason silently, using internal vectors called synthetic continuous CoT (CCoT). And for harder problems—where silence isn’t enough—it smartly reroutes the model back into verbal reasoning mode. This hybrid, adaptive strategy achieves the best of both worlds. ...

For centuries, reading has meant scrolling—page by page, line by line. But what if reading could mean navigating a tree? TreeReader, a new system from researchers at the University of Toronto and the Vector Institute, challenges the linearity of academic literature. It proposes a reimagined interface: one where large language models (LLMs) summarize each section and paragraph into collapsible nodes in a hierarchical tree, letting readers skim, zoom, and verify with surgical precision. The result is more than a UX tweak—it’s a new cognitive model for how scholars might interact with complex documents in the era of AI. ...

What if every story, no matter where it’s set, ends with a cheerful festival and a return to tradition? That’s not a hypothetical. It’s what happens when you ask OpenAI’s GPT-4o-mini to generate 11,800 stories, one for nearly every nationality on Earth. Researchers Jill Walker Rettberg and Hermann Wigers did just that — and uncovered a startling truth: generative AI doesn’t just reproduce representational bias (like stereotyping a “doctor” as a white man), it also imposes narrative bias — structural sameness beneath a veneer of cultural difference. ...

Modern oncology is an overwhelming cognitive battlefield: clinicians face decades of fragmented notes, tests, and treatment episodes, scattered across multiple languages and formats. Large Language Models (LLMs) promise relief—but without careful design, they often collapse under the weight of these chaotic Electronic Health Records (EHRs), hallucinate unsafe recommendations, or fail to reason over time. Enter CliCARE: a meticulously designed framework that not only tames this complexity but grounds the entire decision process in clinical guidelines. Rather than stuffing raw records into long-context transformers or bolting on retrieval-augmented generation (RAG), CliCARE introduces a radically more structured approach. ...

Retrieval-Augmented Generation (RAG) is often hailed as a cure-all for domain adaptation and factual accuracy in large language models (LLMs). By injecting external context at inference time, RAG systems promise to boost performance on knowledge-intensive tasks. But a new paper, RAG in the Wild (Xu et al., 2025), reveals that this promise is brittle when we leave the sanitized lab environment and enter the real world of messy, multi-source knowledge. ...

In quantitative finance, sparse portfolio optimization is a famously unforgiving problem. Selecting the top m assets from a universe of n under budget and risk constraints is NP-hard, highly sensitive to hyperparameters, and often brittle in volatile markets. Traditional solutions—from greedy algorithms to convex relaxations—either crumble under market shifts or produce opaque, overfitted outputs. But what if we reframed the problem entirely? Enter EFS (Evolutionary Factor Search), a radical new framework that turns sparse portfolio construction into an LLM-guided ranking game. Instead of laboriously tuning machine learning models or relying on rigid heuristics, EFS lets large language models generate, evolve, and select alpha factors—and it does so in a way that is not just automated, but interpretable, adaptive, and surprisingly effective. ...

Financial markets don’t reward the loudest opinions. They reward the most timely, well-calibrated ones. FinDPO, a new framework by researchers from Imperial College London, takes this lesson seriously. It proposes a bold shift in how we train language models to read market sentiment. Rather than relying on traditional supervised fine-tuning (SFT), FinDPO uses Direct Preference Optimization (DPO) to align a large language model with how a human trader might weigh sentiment signals in context. And the results are not just academic — they translate into real money. ...

In the age of Digital Twins and hyper-automated warehouses, simulations are everywhere—but insights are not. Discrete Event Simulations (DES) generate rich, micro-level data on logistics flows, delays, and resource utilization, yet interpreting these data remains painfully manual, fragile, and siloed. This paper from Quantiphi introduces a compelling solution: transforming raw simulation outputs into a Knowledge Graph (KG) and querying it via an LLM agent that mimics human investigative reasoning. It’s a shift from spreadsheet-style summaries to an interactive AI assistant that explains why something is slow, where the bottleneck is, and what needs attention. ...

Imagine asking, “Why wasn’t Order A scheduled for production yesterday?” and getting not just an answer, but a causal breakdown, an alternative plan, and a visual comparison — all without involving your operations research (OR) consultant. That’s exactly what SMARTAPS delivers. Built by Huawei researchers, SMARTAPS is a tool-augmented LLM interface for interacting with Advanced Planning Systems (APS) using natural language. It doesn’t try to replace optimization solvers — it simply makes them accessible. In doing so, it redefines how planners interact with complex decision-making models. ...

In the age of natural language interfaces to databases (NLIDBs), asking the right question has never been easier—or more perilous. While systems like ChatGPT or SQL-Palm can convert everyday English into valid SQL, they often do so without interrogating the quality of the question itself. And as Peter Drucker warned, “The most dangerous thing is asking the wrong question.” Enter VeriMinder, a system built not to improve SQL syntax or execution accuracy—but to diagnose and refine the analytical intent behind the user’s query. It tackles a deceptively simple yet far-reaching problem: a well-formed SQL query that answers a poorly formed question can yield confident but misleading insights. This is particularly problematic in enterprise settings where non-technical users rely on LLM-based BI assistants. ...