
MAESTRO Brings Private Deep-Research Automation to Your own Hardware
An open-source, self-hosted research platform coordinates five AI agents to plan, search, reflect, and write reports over your own document library.
20 stories

An open-source, self-hosted research platform coordinates five AI agents to plan, search, reflect, and write reports over your own document library.

Letta's open-source plugin runs a background agent that watches Claude Code sessions, remembers everything, and whispers guidance before your next prompt.

Researchers from Technion, IBM Research, and NVIDIA show that flipping just two bits can reduce ResNet-50 accuracy by 99.8% and collapse a 30B MoE reasoning model.

New method estimates the unknown floor of diffusion training loss, exposing where models still underfit and improving FID by up to 25%.

A new benchmark tests whether LLM agents can decode vague human instructions in a shared task, and top models match human performance.

A new framework decides what LLM agents should remember based on prediction error rather than hand-crafted heuristics, cutting tokens 88% while beating full-context accuracy.

A self-paced curriculum framework matches state-of-the-art reasoning fine-tuning using up to 100x fewer training samples by treating data clusters as bandit arms.

A Microsoft Research and UIUC team shows that random dictionary words can push LLMs past their bias toward LRU and first-fit, producing genuinely novel system algorithms.

A Tsinghua team's DeepPrune framework cuts parallel reasoning token costs by 65 to 88 percent by killing redundant chain-of-thought traces before they finish.

A new retrieval-augmented framework teaches multimodal LLMs to spot the specific ways they usually misread memes, lifting F1 scores by 8.30% on average.

A new paper accepted at L4DC couples reinforcement learning with sampling-based MPC, boosting success rates up to 72% while cutting training time.

New theory pins down exactly how depth, precision, and chain-of-thought shape what state-space models like Mamba can and cannot compute.

A study of 4,514 omics papers finds under 3% report ancestry, threatening biomedical foundation models with cascading population bias.

A new framework lifts reference images into 3D meshes and renders them from four angles to keep inserted objects consistent as cameras move.

A frequency-domain augmentation method reduces sepsis regression error by 10.2% MSE on MIMIC-IV, holding up under shorter observation windows and longer prediction horizons.

A new RL framework has heterogeneous agents collaborate inside a shared search tree, lifting Qwen3-8B by 8 points on LiveCodeBench.

A new graph-based recommender for Steam blends LLM reasoning with signed edge weights to fix the accuracy-vs-diversity trade-off in game suggestions.

A new shared-log abstraction lets AI agents fork production data streams cheaply, isolating their exploratory reads and risky writes from live workloads.

A new framework uses causal inference to pinpoint the specific attention heads driving toxic output, then surgically steers or retrains only those components.

Google Research and DeepMind built a multi-agent Gemini system that mines wearable data for candidate biomarkers, validated across 9,279 participants.