DiffusionGemma is an experimental open-weight language model that generates text through discrete diffusion, refining 256-token blocks in parallel rather than producing tokens sequentially. The design is intended to avoid the decoding bottleneck of autoregressive language models. The model was created by fine-tuning the mixture-of-experts Gemma 4 model, which has 3.8 billion activated parameters and 25.2 billion total parameters. Its two-stage training process used fewer than 10% of the original autoregressive model’s total training-token budget. Supervised fine-tuning first taught bidirectional denoising; a second phase combined reinforcement learning and sampler distillation to improve generation quality and inference efficiency. Across its evaluation suite, DiffusionGemma generated about 20 tokens per forward pass and about 1,500 output tokens per second on a single Nvidia H100 GPU. The researchers say these results establish a new speed-capability trade-off frontier and exceed autoregressive models, including those using speculative decoding. DiffusionGemma retains Gemma 4’s thinking mode, multimodal-input, and long-context support. It can also still generate text autoregressively with minor performance degradation, suggesting potential hybrid diffusion-autoregressive decoding systems.

Researchers from NVIDIA and Flower Labs have developed a method for recursive self-improving AI agents that allows them to enhance their own code without reaching an evaluation ceiling. This advancement addresses a significant challenge in the development of self-improving AI.

Semaglutide is associated with a 26% reduction in the predicted risk of developing dementia over five years. This finding suggests potential cognitive benefits of the drug beyond its use for weight management and diabetes.

A program is proposed to enable a single cell to build itself into a brain using only genomic information. The program must be compact enough to fit within a genome and efficient enough to complete development in a timely manner.

The Material Discovery Bench evaluates models on their ability to propose novel, BEOL-compatible crystalline materials that meet specific targets for thermal conductivity ($kappa), static dielectric constant ($epsilon), Young's modulus, and shear modulus. The benchmark utilizes a harness, tools, and graders to measure model performance based on these criteria.

Research investigates the ability of large language models to introspect on their internal states. The study employs injected representations of known concepts into model activations to differentiate genuine introspection from confabulations.

8 days ago, while jogging, I asked Claude to solve the Riemann Hypothesis It didn’t. 1.5 days later, it proved >= 67% of the zeros are on the line (prev: 41.6%) Still not sure what that means, but some analytic number theorists seem excited - I’d tried once before in a single session and it went nowhere. Eight days ago, when I tried again, this was the opening message:The heavy lifting was recent ...

Claude attempted to address the Riemann hypothesis, a famous unsolved problem in mathematics dating back to 1859. The challenge reflects the ongoing exploration of Claude's mathematical capabilities and its potential contributions to scientific discovery.

AI benchmarks are essential for assessing model performance and informing deployment choices. However, they often reach saturation, reducing their effectiveness in distinguishing between models and limiting their long-term utility.

ChatGPT for Academic Researchers provides 100,000 scientists and mathematicians with free access to advanced ChatGPT models. An AI-generated disproof of the Erdős unit-distance conjecture was shared in May.