Large Language Models (LLMs) often produce hallucinations in retrieval-augmented or long-context generation, even when relevant evidence is present. This stems from two issues: head importance is treated as input-agnostic, and raw attention weights poorly reflect each token's true contribution. We present HAVE (Head-Adaptive Gating and ValuE Calibration), a parameter-free decoding framework that directly addresses both challenges. HAVE introduces head-adaptive gating, which performs instance-level soft reweighing of attention heads, and value calibration, which augments attention with the magnitude of value vectors to approximate write-back contribution. Together, these modules construct token-level evidence aligned with model updates and fuse it with the LM distribution through a lightweight uncertainty-scaled policy. HAVE requires no finetuning and operates in a single forward pass, making it efficient and broadly applicable. Experiments across multiple QA benchmarks and LLM families demonstrate that HAVE consistently reduces hallucinations and outperforms strong baselines, including DAGCD, with modest overhead. The framework is transparent, reproducible, and readily integrates with off-the-shelf LLMs, advancing trustworthy generation in real-world settings.

Large Language Models (LLMs) have demonstrated remarkable emergent capabilities, yet the robustness of their numerical reasoning remains an open question. While standard benchmarks evaluate LLM reasoning on complex problem sets using aggregated metrics, they often obscure foundational weaknesses. In this work, we probe LLM mathematical numeracy by evaluating performance on problems of escalating complexity, from constituent operations to combinatorial puzzles. We test several state-of-the-art LLM-based agents on a 100-problem challenge comprising four categories: (1) basic arithmetic, (2) advanced operations, (3) primality checking, and (4) the Game of 24 number puzzle. Our results show that while the agents achieved high accuracy on the first three categories, which require deterministic algorithmic execution, they consistently failed at the number puzzle, underlining its demand for a heuristic search over a large combinatorial space to be a significant bottleneck. These findings reveal that the agents' proficiency is largely confined to recalling and executing known algorithms, rather than performing generative problem-solving. This suggests their apparent numerical reasoning is more akin to sophisticated pattern-matching than flexible, analytical thought, limiting their potential for tasks that require novel or creative numerical insights.

O1/R1 style large reasoning models (LRMs) signal a substantial leap forward over conventional instruction-following LLMs. By applying test-time scaling to generate extended reasoning paths, they establish many SOTAs across a wide range of complex reasoning tasks. However, recent studies show that LRMs are prone to suffer from overthinking -- the tendency to overcomplicate simple problems, leading to excessive strategy switching and long, convoluted reasoning traces that hinder their interpretability. To mitigate this issue, we conduct a systematic investigation into the reasoning efficiency of a broad set of LRMs and uncover a common dilemma: the difficulty in balancing multiple generation objectives such as correctness and brevity. Based on this discovery, we propose a test-time scaling method, EDIT (Efficient Dynamic Inference Trimming), which efficiently guides LRMs to identify the shortest correct reasoning paths at test time. EDIT employs constraint-guided generation while jointly tracking length and answer distributions under varying constraints, allowing it to select responses that strike an optimal balance between conciseness and correctness. Extensive experiments across diverse models and datasets show that EDIT substantially enhance the reasoning efficiency, producing compact yet informative outputs that improve readability and user experience.

In the field of software traceability link recovery (TLR), textual similarity has long been regarded as the core criterion. However, in tasks involving natural language and programming language (NL-PL) artifacts, relying solely on textual similarity is limited by their semantic gap. To this end, we conducted a large-scale empirical evaluation across various types of TLR tasks, revealing the limitations of textual similarity in NL-PL scenarios. To address these limitations, we propose an approach that incorporates multiple domain-specific auxiliary strategies, identified through empirical analysis, into two models: the Heterogeneous Graph Transformer (HGT) via edge types and the prompt-based Gemini 2.5 Pro via additional input information. We then evaluated our approach using the widely studied requirements-to-code TLR task, a representative case of NL-PL TLR. Experimental results show that both the multi-strategy HGT and Gemini 2.5 Pro models outperformed their original counterparts without strategy integration. Furthermore, compared to the current state-of-the-art method HGNNLink, the multi-strategy HGT and Gemini 2.5 Pro models achieved average F1-score improvements of 3.68% and 8.84%, respectively, across twelve open-source projects, demonstrating the effectiveness of multi-strategy integration in enhancing overall model performance for the requirements-code TLR task.

This research addresses a fundamental question in AI: whether large language models truly understand concepts or simply recognize patterns. The authors propose bidirectional reasoning,the ability to apply transformations in both directions without being explicitly trained on the reverse direction, as a test for genuine understanding. They argue that true comprehension should naturally allow reversibility. For example, a model that can change a variable name like userIndex to i should also be able to infer that i represents a user index without reverse training. The researchers tested current language models and discovered what they term cognitive specialization: when models are fine-tuned on forward tasks, their performance on those tasks improves, but their ability to reason bidirectionally becomes significantly worse. To address this issue, they developed Contrastive Fine-Tuning (CFT), which trains models using three types of examples: positive examples that maintain semantic meaning, negative examples with different semantics, and forward-direction obfuscation examples. This approach aims to develop deeper understanding rather than surface-level pattern recognition and allows reverse capabilities to develop naturally without explicit reverse training. Their experiments demonstrated that CFT successfully achieved bidirectional reasoning, enabling strong reverse performance while maintaining forward task capabilities. The authors conclude that bidirectional reasoning serves both as a theoretical framework for assessing genuine understanding and as a practical training approach for developing more capable AI systems.

Large Language Models (LLMs) are increasingly used for recommendation tasks due to their general-purpose capabilities. While LLMs perform well in rich-context settings, their behavior in cold-start scenarios, where only limited signals such as age, gender, or language are available, raises fairness concerns because they may rely on societal biases encoded during pretraining. We introduce a benchmark specifically designed to evaluate fairness in zero-context recommendation. Our modular pipeline supports configurable recommendation domains and sensitive attributes, enabling systematic and flexible audits of any open-source LLM. Through evaluations of state-of-the-art models (Gemma 3 and Llama 3.2), we uncover consistent biases across recommendation domains (music, movies, and colleges) including gendered and cultural stereotypes. We also reveal a non-linear relationship between model size and fairness, highlighting the need for nuanced analysis.

No, AI isn't going to kill us all, despite what this new book says In the totality of human existence, there are an awful lot of things for us to worry about. Money troubles, climate change and finding love and happiness rank highly on the list for many people, but for a dedicated few, one concern rises above all else: that artificial intelligence will eventually destroy the human race. Eliezer Yudkowsky at the Machine Intelligence Research Institute (MIRI) in California has been proselytising this cause for a quarter of a century, to a small if dedicated following. Then we entered the ChatGPT era, and his ideas on AI safety were thrust into the mainstream, echoed by tech CEOs and politicians alike. Writing with Nate Soares, also at MIRI, is Yudkowsky's attempt to distil his argument into a simple, easily digestible message that will be picked up across society.

The world is full of extraordinary young people brimming with ideas for how to crack tough problems. Every year, we recognize 35 such individuals from around the world--all of whom are under the age of 35. These scientists, inventors, and entrepreneurs are working to help mitigate climate change, accelerate scientific progress, and alleviate human suffering from disease. Some are launching companies while others are hard at work in academic labs. They were selected from hundreds of nominees by expert judges and our newsroom staff. Get to know them all--including our 2025 Innovator of the Year-- in these profiles .

How Yichao "Peak" Ji became a global AI app hitmaker He developed Manus, one of the buzziest AI apps of the year, in the latest project that blends his technical prowess with killer consumer instincts. When Yichao Ji--also known as "Peak"--appeared in a launch video for Manus in March, he didn't expect it to go viral. Speaking in fluent English, the 32-year-old introduced the AI agent built by Chinese startup Butterfly Effect, where he serves as chief scientist. The video was not an elaborate production--it was directed by cofounder Zhang Tao and filmed in a corner of their Beijing office. But something about Ji's delivery, and the vision behind the product, cut through the noise. The product, then still an early preview available only through invite codes, spread across the Chinese internet to the world in a matter of days.

The chatbot-enabled Friend necklace eavesdrops on your life and provides a running commentary that's snarky and unhelpful. Worse, it can also make the people around you uneasy. The AI-powered Friend pendant is now out in the world. If you live in the US or Canada, you can buy one for $129. The smooth plastic disc is just under 2 inches in diameter; it looks and feels a little like a beefy Apple AirTag. Inside are some LEDs and a Bluetooth radio that connects you (through your iPhone) to a chatbot in the cloud that's powered by Google's Gemini 2.5 model. You can tap on the disc to ask your Friend questions as it dangles around your neck, and it responds to your voice prompts by sending you text messages through the companion app.