A recursive logic with five principles is proposed to improve mechanistic inference in interpretability.
What is new.
What it means for your work.
Discover recent technology and evidence. Browse older material in All notes.
Changes to start with
Newly added notes across different industries
AgentRoboCritic · Action-Level Analysis for Evaluating Robot Control Capabilities
TideCache: Transparent Cache Management and Layer-wise Streaming for Hybrid LLM Serving
Virus-like particles enable efficient gene editing in pigs and poultry models
New technology notes
Originals from the past 14 days, ordered by when added to Blackdeck.
A game-based framework is introduced to explore how LLMs discover research ideas through yes/no questions.
A single representation space exhibits emergent capabilities at functional, residue, and bond levels
Custom instructions enable accelerator orchestration and reduced load/store latency for AI workloads.
Introduces a bit-plane-selective, sensitivity-guided scheme to protect AI weights without extra storage overhead
Scientific hypotheses are replaced with sets of mutually exclusive and exhaustive competing explanations to clarify interpretation
Supports differentiation through both inputs and curve coefficients, enabling differentiable parametric curve learning.
A mechanism by which exoplanets in binary white dwarf systems imprint their presence on gravitational wave signals is proposed
A single VLA policy handles all 100 long-horizon household tasks efficiently.
A solid-state battery using sulfide electrolyte completes 1,000 full cycles at 80% capacity.
Learning experience is analyzed by AI to generate adaptive, self-adjusting scaffolds during training.
Integrates diverse medical imaging datasets with anatomical and abnormal annotations
A range of adaptation strategies for DINOv2 foundation model is tested across medical imaging tasks
Introduces region-specific correction for panel-level pattern distortion, enabling precise overlay alignment
Same clinical scenarios yield different intervention recommendations depending on whether the input is from clinician, caregiver, or patient