AI Agents
AI Agent technology and applications
Articles
Google DeepMind Unveils AlphaGenome Atlas: Precomputed Molecular⭐6
Google DeepMind’s AlphaGenome Atlas predicts molecular effects of 9B human DNA variants with AVI scores, now free for academic research.
Reducto Releases r-1: A Single-Pass Document Parsing Model That⭐8
Reducto's r-1 model cuts parsing errors by 20% with a single-pass design, offering low-cost document parsing at one cent per page.
OpenBMB Releases MiniCPM5-2B: A 2.52B Dense Model Averaging 53.9⭐7
OpenBMB's MiniCPM5-2B: a 2.52B dense model scoring 53.9 across 34 benchmarks, excelling in code and tool use for on-device deployment.
Axis Robotics Unveils AXIS: A Browser-Based Data Engine⭐8
AXIS introduces a browser-based data engine with 207 robot tasks and 50,129 trajectories, transforming scalable, open robot data collection.
IFM Unveils K2 Horizon: Six Apache 2.0 Models from 0.9B to 375B⭐8
IFM launches K2 Horizon: six Apache 2.0 models from 0.9B to 375B-A23B with open weights, training data, and code for fully transparent AI development.
Meta FAIR Introduces AI Research Preference Models (RPMs):⭐7
Meta FAIR's new RPMs rank ML experiments before GPU-hours are spent, cutting wasted compute and improving research agent efficiency.
H Company Launches NeoMME: A 260M and 800M Single-Tower⭐7
H Company's NeoMME delivers 260M and 800M single-tower multimodal encoders, removing vision towers and causal decoders for faster, efficient document retrieval.
UC Berkeley Researchers Release CUA-Lite, an Open Platform⭐8
UC Berkeley’s CUA-Lite unifies sandboxes, data, evaluation, and RL into an open platform for advancing computer-use AI agents.
Perplexity Details Its GPU Embedding Stack: How Ivy, Tulip, and⭐8
Perplexity details its GPU embedding stack, revealing how Ivy, Tulip, and ROSE power efficient pplx-embed serving and ranking across search and API platforms.
Nous Research Adds One-Click Local Model Setup to Hermes Desktop⭐8
Nous Research's Hermes Desktop now offers one-click local model setup, auto-detecting hardware and configuring AI models, making local deployment effortless.