Evaluation of 3D medical foundation models is typically dominated by downstream task scores, which entangle representation quality with
Source excerpt
Methods worth understanding
New AI methods, the evidence behind them, and how to put them to work.
Latest weekly report
Methods worth understanding, their results, and practical limitations.
Evaluation of 3D medical foundation models is typically dominated by downstream task scores, which entangle representation quality with
Source excerpt
Align a source robot’s visual encoder to a frozen target robot’s visual encoder by mapping paired source–target observations into a shared latent space, eliminating an embodiment-induced observation gap. A policy trained on encoded source demonstrations can then act using target-robot observations directly, without pixel-level image translation.
AI summary
Use a dedicated research platform that pairs a child-friendly conversational interface with configurable age-based safety guardrails, researcher oversight, and privacy-preserving data handling to enable controlled study of how minors interact with LLMs.
AI summary
Selected-Risk Calibration (SRC) is a framework for calibrating a learned surrogate constraint using only the optimizer’s returned decisions. Instead of guaranteeing coverage over random inputs, SRC targets the probability that a decision produced by the surrogate-based optimizer is actually infeasible under the true constraint, for objectives drawn from a specified distribution.
AI summary
During multi-request, multi-LoRA inference, release each request’s LoRA adapter before generation completes and continue decoding using only the shared backbone. Pair this with GPU kernels that are aware of which batch rows are still using LoRA, and dynamically resize kernel launches so idle compute blocks are trimmed as adapters are released.
AI summary
Compare model behavior associated with sycophancy across training checkpoints from different pipeline stages (e.g., supervised fine-tuning vs preference alignment) and use representation-level probes to link sycophancy to specific training data sources within each stage.
AI summary
Train a small “student” segmentation network by distilling knowledge from a larger “teacher” model. The approach targets label-scarce settings by using the teacher’s outputs to guide the student, aiming to retain segmentation quality while reducing model complexity.
AI summary
Use a grouped, multi-directional variant of Mamba inside an encoder for 3D segmentation: split input channels into G groups, flatten each group in multiple scan directions within one pass to aggregate multi-directional context, and apply Mamba only at deeper encoder stages to keep the model smaller.
AI summary
Use a shared experimentation framework that enforces fixed data and compute budgets per run while keeping the downstream evaluation protocol constant, and record all runs in a public experiment ledger for others (including AI agents) to search, reproduce, and build on prior attempts.
AI summary
The technique is to benchmark representation quality for 3D brain MRI by evaluating a collection of pre-trained vision encoders on shared downstream tasks. It compares encoders pretrained on natural 2D RGB data (for which domain transfer may be imperfect) against encoders pretrained on medical imaging, including both supervised and self-supervised variants, while holding the evaluation protocol consistent across models.
AI summary
Large Language Models (LLMs) are emerging as natural-language interfaces for Building Information Modeling (BIM), successfully achieving information retrieval,
Source excerpt
Use a Stable-Diffusion-like diffusion model where selected feed-forward layers and/or attention layers are replaced with a sparse mixture-of-experts (MoE) block. An internal router network selects a subset of expert networks to process each token, while the non-selected parts of the network remain shared with the base model.
AI summary