11 articles — 🟡 4 important , 🟢 7 interesting
DiffusionGemma is Google's diffusion language model operating in continuous latent space. A study by 13 authors led by Neel Nanda shows that the initial opacity is 28.6× greater than Gemma 4, but an interpretable token bottleneck narrows that gap to just 1.1×.
SSD (Spatially Speculative Decoding) is a new method that simultaneously predicts the horizontal and vertical neighbor of a pixel in autoregressive image generation, achieving up to 13.3× speedup without any loss of visual quality on the DPG-Bench and GenEval benchmarks.
TimeProVe is a framework that accelerates VLM inference over long videos by introducing a two-stage propose-then-verify approach. It reduces calls to expensive models by 75% and total inference cost by 93%, while outperforming the strongest competitor by 7.3 percentage points on the new OpenTSUBench benchmark.
VIMPO is a new reinforcement learning method for LLM reasoning that derives an implicit value function from KL-regularized RL — without a separate critic network. It outperforms GRPO on four mathematical benchmarks including AIME 2024 and AIME 2025, with gains that remain stable even under noisy reward conditions.
HumanScale is a systematic comparison (Peking University and MIT, 21 authors) showing that filtered human egocentric video yields 52.5 % higher success on familiar tasks and up to 90 % higher success on out-of-distribution robot manipulation tasks compared to models pre-trained exclusively on robot data.
Contagion Networks is a mathematical framework that quantifies the spread of bias between LLM agents. In experiments with three DeepSeek-chat agents, contagion coefficients γ range from 0.157 to 0.352, while expanding the evaluator panel from 1 to 3 members reduces contagion by 72.4%.
H-RePlan is a system that enables AI agents managing Linux and Android simultaneously to autonomously recover from errors — each platform's local recovery strategy is separated from the global orchestrator's replanning, reducing token costs and increasing task success rates.
Abdul Rafay Syed identified a common direction in the activation space of four LLM families — Qwen2.5, Gemma-2, Llama-3.2, and Ministral-3 — that separates aligned from misaligned models with 99.6% accuracy, while directional steering reduces unsafe code leakage by 21–51 points.
NeuroImprint is an attack that corrupts PEFT adapters in federated fine-tuning and reconstructs 59–79% of all training samples with high semantic fidelity. Tested on BERT, GPT-2, Qwen2, and Llama 3.2, the attack remains undetected because the model retains normal utility.
