🔍 Mechanistic Interpretability - ghosh.debasish · Scour

Automated Attention Pattern Discovery at Scale in Large Language Models 🧠LLMs

arxiv.org·3d

Scientists Just Witnessed Superposition In Atoms For the Very First Time 📜Operational Semantics

popularmechanics.com·3d

Latent Reasoning Sprint #3: Activation Difference Steering and Logit Lens 🧠Reasoning Models

lesswrong.com·6d

XENONOSTRA RESEARCH NOTES Topological Emotion Vector Monitoring for Real-Time LLM Alignment: A… 🔁Category Theory

medium.com·6d

Disentangling MLP Neuron Weights in Vocabulary Space 🧠LLMs

arxiv.org·2d

US physicists directly observe atomic wave superposition 📜Operational Semantics

kite.kagi.com·3d

Information as Structural Alignment: A Dynamical Theory of Continual Learning 🎯RLHF

arxiv.org·1d

HAWK: Head Importance-Aware Visual Token Pruning in Multimodal Models 🎲Probabilistic Data Structures

arxiv.org·9h

Spectral Edge Dynamics Reveal Functional Modes of Learning 🎯RLHF

arxiv.org·1d

ConceptTracer: Interactive Analysis of Concept Saliency and Selectivity in Neural Representations 🗄️Vector Databases

arxiv.org·1d

The Illusion of Superposition? A Principled Analysis of Latent Thinking in Language Models 🧠Reasoning Models

arxiv.org·1d

Improving Robustness In Sparse Autoencoders via Masked Regularization 🧠LLMs

arxiv.org·1d

Selective Aggregation of Attention Maps Improves Diffusion-Based Visual Interpretation 🌊Streaming Algorithms

arxiv.org·2d

AudioKV: KV Cache Eviction in Efficient Large Audio Language Models 🧠LLMs

arxiv.org·1d

Geometric Limits of Knowledge Distillation: A Minimum-Width Theorem via Superposition Theory 🧠LLMs

arxiv.org·3d

SMT-AD: a scalable quantum-inspired anomaly detection approach 🧠Reasoning Models

arxiv.org·1d

Out-of-Air Computation: Enabling Structured Extraction from Wireless Superposition 🌊Streaming Algorithms

arxiv.org·3d

Design and Analysis of Chirp-Layered Superposition Coding for LoRa 🌊Streaming Algorithms

arxiv.org·2d

Finding Belief Geometries with Sparse Autoencoders 🧠Reasoning Models

arxiv.org·4d

Isokinetic Flow Matching for Pathwise Straightening of Generative Flows 🌊Streaming Algorithms

arxiv.org·3d

Loading more...