Challenging the prevailing consensus that small models inherently lack robust reasoning, this report introduces VibeThinker-1.5B, a 1.5B-parameter dense model developed via our Spectrum-to-Signal Principle (SSP). This challenges the prevailing approach of scaling model parameters to enhance capabilities, as seen in models like DeepSeek R1 (671B) and Kimi k2 (>1T). The SSP framework first employs a Two-Stage Diversity-Exploring Distillation (SFT) to generate a broad spectrum of solutions, followed by MaxEnt-Guided Policy Optimization (RL) to amplify the correct signal. With a total training cost of only $7,800, VibeThinker-1.5B demonstrates superior reasoning capabilities compared to closed-source models like Magistral Medium and Claude Opus 4, and performs on par with open-source models like GPT OSS-20B Medium. Remarkably, it surpasses the 400x larger DeepSeek R1 on three math benchmarks: AIME24 (80.3 vs. 79.8), AIME25 (74.4 vs. 70.0), and HMMT25 (50.4 vs. 41.7). This is a substantial improvement over its base model (6.7, 4.3, and 0.6, respectively). On LiveCodeBench V6, it scores 51.1, outperforming Magistral Medium's 50.3 and its base model's 0.0. These findings demonstrate that small models can achieve reasoning capabilities comparable to large models, drastically reducing training and inference costs and thereby democratizing advanced AI research.

The design of novel molecules with desired properties is a key challenge in drug discovery and materials science. Traditional methods rely on trial-and-error, while recent deep learning approaches have accelerated molecular generation. However, existing models struggle with generating molecules based on specific textual descriptions. We introduce Mol-CADiff, a novel diffusion-based framework that uses causal attention mechanisms for text-conditional molecular generation. Our approach explicitly models the causal relationship between textual prompts and molecular structures, overcoming key limitations in existing methods. We enhance dependency modeling both within and across modalities, enabling precise control over the generation process. Our extensive experiments demonstrate that Mol-CADiff outperforms state-of-the-art methods in generating diverse, novel, and chemically valid molecules, with better alignment to specified properties, enabling more intuitive language-driven molecular design.

Multimodal learning is a rapidly growing research field that has revolutionized multitasking and generative modeling in AI. While much of the research has focused on dealing with unstructured data (e.g., language, images, audio, or video), structured data (e.g., tabular data, time series, or signals) has received less attention. However, many industry-relevant use cases involve or can be benefited from both types of data. In this work, we propose a modular, end-to-end multimodal learning method called MAGNUM, which can natively handle both structured and unstructured data. MAGNUM is flexible enough to employ any specialized unimodal module to extract, compress, and fuse information from all available modalities.

It may not be obvious how internet service providers can use AI to boost business. Don't they simply provide the amount and type of internet their customers are willing to buy? In reality, providing internet service is much larger than installing a router.

Machine learning is the hot buzzword of the last couple years. While some applications have been a little strange, it's likely that machine learning can make shopping a little easier by inferring a shopper's interests based on their history with the store. Valve unveiled Steam Labs to share its machine learning results with the world. Two of the Steam Labs experiments revolve around video content. The Micro Trailers experiment employs deep learning to automatically trim official game trailers into six-second video clips.

In extreme classification problems, learning algorithms are required to map instances to labels from an extremely large label set. We build on a recent extreme classification framework with logarithmic time and space, and on a general approach for error correcting output coding (ECOC), and introduce a flexible and efficient approach accompanied by bounds. Our framework employs output codes induced by graphs, and offers a tradeoff between accuracy and model size. We show how to find the sweet spot of this tradeoff using only the training data. Our experimental study demonstrates the validity of our assumptions and claims, and shows the superiority of our method compared with state-of-the-art algorithms.

A set of Python Reinforcement Learning environments powered by the MuJoCo physics engine. Libraries that provide Python bindings to the MuJoCo physics engine. If you use this package, please cite our accompanying accompanying tech report. MuJoCo Pro must be installed before dm_control, since dm_control's install script generates Python ctypes bindings based on MuJoCo's header files. By default, dm_control assumes that the MuJoCo Zip archive is extracted as /.mujoco/mjpro150.

NVIDIA just announced the Jetson TX2 embedded AI supercomputer, based on the latest NVIDIA Pascal microarchitecture. It promises to offer twice the performance of the previous-generation Jetson TX1, in the same package. In this tech report, we will share with you the full details of the new Pascal-based NVIDIA Jetson TX2! Artificial intelligence is the new frontier in GPU compute technology. Whether they are used to power training or inference engines, AI research has benefited greatly from the massive amounts of compute power in modern GPUs. The market is led by NVIDIA with their Tesla accelerators that run on their proprietary CUDA platform.

Despite the rise of machine-learning and other artificial-intelligence tools, it only seems to get harder and harder to find just the right gifts for the nerd in your life. We on the TR staff know just how hard to both pick gifts for our favorite techies and to have gifts picked out for us. That's why we're continuing our annual tradition of compiling the unusual, the useful, and the delightful items that we've used in the past year. Whatever your budget, we've got something for the nerd in your life. Our gift guide is sponsored by Newegg.