From Local Recovery to Unified Generation: Discovering Triple Set in Knowledge Graphs via Observation-Anchored Latent Diffusion introduces LatentDiff, a unified latent generative framework for triple set prediction. It encodes the whole graph into relation-aware latent embeddings with a residual-quantized autoencoder and uses an observation-anchored latent diffusion model for denoising. Experiments on three benchmarks show it outperforms the state-of-the-art by 5.4 percentage points in Fₜₛₚ.
Key points
LatentDiff uses a relation-aware residual-quantized knowledge graph autoencoder.
It improves Fₜₛₚ by 5.4 percentage points over the previous state-of-the-art on three benchmark datasets.
Sources
Self-Supervised Doppler-Guided RF Odometry accepted at NeurIPS 2026
Self-Supervised Doppler-Guided RF Odometry is a new RF-only odometry framework called RFWalk, accepted at NeurIPS 2026. It learns dense correspondences via random walks and a Doppler consistency loss, and reduces absolute trajectory error by 58% in dynamic scenes with moving pedestrians.
Key points
RFWalk models dense correspondences across range-azimuth frames as random walks on a space-time graph.
RFWalk reduces absolute trajectory error by 58% in dynamic scenes with moving pedestrians.
Sources
- Original post
- Self-Supervised Doppler-Guided RF Odometry - Linked page
Meta signs 20-year deal with Constellation, adds 30 MW
Meta signed a 20-year agreement with Constellation Energy to support an Illinois nuclear plant. The deal includes increasing the plant’s output by 30 megawatts.
Key points
Agreement: 20-year agreement with Constellation Energy.
Output increase: plant output up 30 megawatts.
Sources
Open Robotics build farm handles Rust dependencies
Open Robotics build farm can handle Rust dependencies in ROS packages. Thanks to support from OSRA and a lot of work from OSRA members @ekumenlabs and @IntrinsicAI.
Key points
Build farm capability: can handle Rust dependencies in ROS packages
Contributors: OSRA, @ekumenlabs, @IntrinsicAI
Sources
- Original post
- osralliance.org - Linked in the post
Seminar - Stephanie Gil on Trust, Sensing, Learning
Seminar - Stephanie Gil on Trust, Sensing, Learning announces a talk at ETH Zurich on 6 Nov 2026 at 16:00. All welcome.
Key points
Date & time: Fri 6 Nov 2026, 16:00
Location: ETH Zurich, HG G 3
Sources
Lumentum and Marvell partner on ultra-low latency optical switches
Lumentum and Marvell partner to supply interconnects for high-density compute fabrics, launching a multi-year effort that equips hyperscale data centers with ultra-low latency co-packaged optical switches. The initiative targets photonics deployment as copper reaches physical limits.
Key points
Partnership: Lumentum and Marvell will supply interconnects for high-density compute fabrics.
Initiative: Multi-year effort equips hyperscale data centers with ultra-low latency co-packaged optical switches.
