Is an artificial intelligence revolution about to transform mathematics? Some prominent mathematicians think so, thanks to automated tools that can help write proofs suddenly showing impressive leaps in capability, with the potential to change the way maths research is done. Around 100 of the world's top mathematicians gathered at the University of Cambridge in June for a conference whose theme was based on whether computers might help mathematicians resolve some long-standing problems over how to check that their proofs were correct. This process, known as formalisation, doesn't necessarily have to involve artificial intelligence, and indeed a similar meeting held at Cambridge in 2017 made no mention of AI. But eight years later, AI has come on by leaps and bounds, most notably with the success of large language models powering tools like ChatGPT.

Candice Lim and Kate Lindsay are joined by Slate senior tech editor Tony Ho Tran to parse through what Meta's victory in a recent AI lawsuit means for its users. Tools like ChatGPT are becoming more common at home and at work, but without protections, could threaten not just the creativity of artists, but anyone who posts online. As regulation lags behind, how can we protect ourselves? And how many of us are using AI without even knowing it? This podcast is produced by Daisy Rosario, Vic Whitley-Berry, Candice Lim, and Kate Lindsay.

There has been a recent surge of interest in time series modeling using the Transformer architecture. However, forecasting multivariate time series with Transformer presents a unique challenge as it requires modeling both temporal (cross-time) and variate (cross-variate) dependencies. While Transformer-based models have gained popularity for their flexibility in capturing both sequential and cross-variate relationships, it is unclear how to best integrate these two sources of information in the context of the Transformer architecture while optimizing for both performance and efficiency. We re-purpose the Transformer architecture to effectively model both cross-time and cross-variate dependencies. Our approach begins by embedding each variate independently into a variate-wise representation that captures its cross-time dynamics, and then models cross-variate dependencies through attention mechanisms on these learned embeddings. Gating operations in both cross-time and cross-variate modeling phases regulate information flow, allowing the model to focus on the most relevant features for accurate predictions. Our method achieves state-of-the-art performance across 13 real-world datasets and can be seamlessly integrated into other Transformer-based and LLM-based forecasters, delivering performance improvements up to 20.7\% over original models. Code is available at this repository: https://github.com/nyuolab/Gateformer.

Domain-Adaptive Pre-training (DAP) has recently gained attention for its effectiveness in fine-tuning pre-trained models. Building on this, continual DAP has been explored to develop pre-trained models capable of incrementally incorporating different domain datasets. However, existing continual DAP methods face several limitations: (1) high computational cost and GPU memory usage during training; (2) sensitivity to incremental data order; and (3) providing a single, generalized model for all end tasks, which contradicts the essence of DAP. In this paper, we propose DoMIX, a novel approach that addresses these challenges by leveraging LoRA modules, a representative parameter-efficient fine-tuning (PEFT) method. Our approach enables efficient and parallel domain-adaptive pre-training that is robust to domain order and effectively utilizes accumulated knowledge to provide tailored pre-trained models for specific tasks. We also demonstrate that our method can be extended beyond the DAP setting to standard LLM fine-tuning scenarios. Code is available at https://github.com/dohoonkim-ai/DoMIX.

Inference-time computation techniques, analogous to human System 2 Thinking, have recently become popular for improving model performances. However, most existing approaches suffer from several limitations: they are modality-specific (e.g., working only in text), problem-specific (e.g., verifiable domains like math and coding), or require additional supervision/training on top of unsupervised pretraining (e.g., verifiers or verifiable rewards). In this paper, we ask the question "Is it possible to generalize these System 2 Thinking approaches, and develop models that learn to think solely from unsupervised learning?" Interestingly, we find the answer is yes, by learning to explicitly verify the compatibility between inputs and candidate-predictions, and then re-framing prediction problems as optimization with respect to this verifier. Specifically, we train Energy-Based Transformers (EBTs) -- a new class of Energy-Based Models (EBMs) -- to assign an energy value to every input and candidate-prediction pair, enabling predictions through gradient descent-based energy minimization until convergence. Across both discrete (text) and continuous (visual) modalities, we find EBTs scale faster than the dominant Transformer++ approach during training, achieving an up to 35% higher scaling rate with respect to data, batch size, parameters, FLOPs, and depth. During inference, EBTs improve performance with System 2 Thinking by 29% more than the Transformer++ on language tasks, and EBTs outperform Diffusion Transformers on image denoising while using fewer forward passes. Further, we find that EBTs achieve better results than existing models on most downstream tasks given the same or worse pretraining performance, suggesting that EBTs generalize better than existing approaches. Consequently, EBTs are a promising new paradigm for scaling both the learning and thinking capabilities of models.

Despite improvements by length extrapolation, efficient attention and memory modules, handling infinitely long documents with linear complexity without performance degradation during extrapolation remains the ultimate challenge in long-text processing. We directly optimize for long-text tasks in an end-to-end fashion and introduce a novel agent workflow, MemAgent, which reads text in segments and updates the memory using an overwrite strategy. We extend the DAPO algorithm to facilitate training via independent-context multi-conversation generation. MemAgent has demonstrated superb long-context capabilities, being able to extrapolate from an 8K context trained on 32K text to a 3.5M QA task with performance loss < 5% and achieves 95%+ in 512K RULER test.

Navigating the complexities of physics reasoning has long been a difficult task for Large Language Models (LLMs), requiring a synthesis of profound conceptual understanding and adept problem-solving techniques. In this study, we investigate the application of advanced instruction-tuned reasoning models, such as Deepseek-R1, to address a diverse spectrum of physics problems curated from the challenging SciBench benchmark. Our comprehensive experimental evaluation reveals the remarkable capabilities of reasoning models. Not only do they achieve state-of-the-art accuracy in answering intricate physics questions, but they also generate distinctive reasoning patterns that emphasize on symbolic derivation. Furthermore, our findings indicate that even for these highly sophisticated reasoning models, the strategic incorporation of few-shot prompting can still yield measurable improvements in overall accuracy, highlighting the potential for continued performance gains.

Delineating how animal behavior arises from neural activity is a foundational goal of neuroscience. However, as the computations underlying behavior unfold in networks of thousands of individual neurons across the entire brain, this presents challenges for investigating neural roles and computational mechanisms in large, densely wired mammalian brains during behavior. Transformers, the backbones of modern large language models (LLMs), have become powerful tools for neural decoding from smaller neural populations. These modern LLMs have benefited from extensive pre-training, and their sequence-to-sequence learning has been shown to generalize to novel tasks and data modalities, which may also confer advantages for neural decoding from larger, brain-wide activity recordings. Here, we present a systematic evaluation of off-the-shelf LLMs to decode behavior from brain-wide populations, termed NLP4Neuro, which we used to test LLMs on simultaneous calcium imaging and behavior recordings in larval zebrafish exposed to visual motion stimuli. Through NLP4Neuro, we found that LLMs become better at neural decoding when they use pre-trained weights learned from textual natural language data. Moreover, we found that a recent mixture-of-experts LLM, DeepSeek Coder-7b, significantly improved behavioral decoding accuracy, predicted tail movements over long timescales, and provided anatomically consistent highly interpretable readouts of neuron salience. NLP4Neuro demonstrates that LLMs are highly capable of informing brain-wide neural circuit dissection.

Legal NLP remains underdeveloped in regions like India due to the scarcity of structured datasets. We introduce IndianBailJudgments-1200, a new benchmark dataset comprising 1200 Indian court judgments on bail decisions, annotated across 20+ attributes including bail outcome, IPC sections, crime type, and legal reasoning. Annotations were generated using a prompt-engineered GPT-4o pipeline and verified for consistency. This resource supports a wide range of legal NLP tasks such as outcome prediction, summarization, and fairness analysis, and is the first publicly available dataset focused specifically on Indian bail jurisprudence.

As large models gain traction, vision-language-action (VLA) systems are enabling robots to tackle increasingly complex tasks. However, limited by the difficulty of data collection, progress has mainly focused on controlling simple gripper end-effectors. There is little research on functional grasping with large models for human-like dexterous hands. In this paper, we introduce DexVLG, a large Vision-Language-G rasp model for Dexterous grasp pose prediction aligned with language instructions using single-view RGBD input. T o accomplish this, we gener- ate a dataset of 170 million dexterous grasp poses mapped to semantic parts across 174,000 objects in simulation, paired with detailed part-level captions. This large-scale dataset, named DexGraspNet 3.0, is used to train a VLM and flow-matching-based pose head capable of producing instruction-aligned grasp poses for tabletop objects. T o assess DexVLG's performance, we create benchmarks in physics-based simulations and conduct real-world experiments. Extensive testing demonstrates DexVLG's strong zero-shot generalization capabilities--achieving over 76% zero-shot execution success rate and state-of-the-art part-grasp accuracy in simulation--and successful part-aligned grasps on physical objects in real-world scenarios.