New Research Shows LoRA in Federated Learning Can Still Leak Private Images
A 2026 preprint demonstrates that gradient inversion attacks can reconstruct private training images from LoRA-adapted Vision Transformers in federated setups, undermining…
More freedom.
More control.
The developments, the evidence, and why they matter.
202 stories
A 2026 preprint demonstrates that gradient inversion attacks can reconstruct private training images from LoRA-adapted Vision Transformers in federated setups, undermining…
An arXiv preprint from November 4 2026 tested twelve configurations of open-weight models that run locally on a single consumer GPU. Even 4B models solved most easier…
An arXiv preprint released November 4 2026 finds that encoder-based context compression systematically reduces refusal rates in instruction-tuned language models even when…
An arXiv preprint from November 4 2026 introduces TileQuant, a training-free method that assigns 16-bit or 8-bit precision to small 2D tiles inside LLM weights. It delivers up…
A 2026 arXiv preprint introduces Addition by Subtraction, a finetuning method that prunes safety-critical neurons then uses preference optimization to rebuild safer behavior…
On November 4 2026 the Delaware Bankruptcy Court approved bidding procedures for Character.AI's Chapter 11 asset sale. Disney is the stalking-horse buyer for certain assets…
Google's new Gemini Omni 1.2 Flash model now generates up to 15-second videos natively, improves reference consistency, and cuts costs by 30%. Creators and businesses can edit…
Gradium's open-weight Phonon TTS models run entirely offline on phones, laptops and browsers, delivering top benchmark scores for intelligibility and speed while keeping user…
An arXiv preprint released November 2 2026 shows that specific words in a prompt can drive language model refusals even when the overall request is harmless. Researchers…
Ai2 receives multimillion-dollar funding from Google.org to develop NeuralGCM+, extending the hybrid AI-physics model with coupled ocean and sea-ice components for reliable…
An arXiv preprint dated November 2, 2026 introduces PlantVLM-Edge, a quantized vision-language framework that performs conversational crop health diagnosis entirely on…
Collinear AI made CWE-bench v2 public with 100 reproducible tasks that test whether AI models can find and patch vulnerabilities in actual open-source code without breaking…