惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

The GitHub Blog
The GitHub Blog
I
InfoQ
U
Unit 42
WordPress大学
WordPress大学
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
Apple Machine Learning Research
Apple Machine Learning Research
J
Java Code Geeks
月光博客
月光博客
D
Docker
Stack Overflow Blog
Stack Overflow Blog
D
DataBreaches.Net
阮一峰的网络日志
阮一峰的网络日志
Blog — PlanetScale
Blog — PlanetScale
V
Visual Studio Blog
博客园 - 聂微东
A
About on SuperTechFans
腾讯CDC
Jina AI
Jina AI
Microsoft Azure Blog
Microsoft Azure Blog
GbyAI
GbyAI
博客园 - 【当耐特】
罗磊的独立博客
博客园 - 三生石上(FineUI控件)
M
MIT News - Artificial intelligence

Goodfire Research

Models know when they’re reward hacking — and we can catch them at scale - Goodfire Forking Fast: Efficiently Estimating Uncertainty Dynamics in Text Generation Steering Along Manifolds to Control Neural Networks Uncovering Neural Geometry in Vision Models With Block-Sparse Featurizers Uncovering Neural Geometry in Vision Models With Block-Sparse Featurizers Meandering on Manifolds: The Neural Geometry of Stories Over Time Can SAEs Capture Neural Geometry? A Geometric Calculator Inside a Neural Network Meandering on Manifolds: The Neural Geometry of Stories Over Time Interpreting Language Model Parameters Predictive Data Debugging: Reveal and Shape What Your Model Learns, Before You Train Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention Predictive Data Debugging: Reveal and Shape What Your Model Learns, Before You Train Logits as a new monitor for evaluation awareness Predicting Rare LLM Failures with 30× Fewer Rollouts Logits as a new monitor for evaluation awareness Predicting Rare LLM Failures with 30× Fewer Rollouts The Shape of Stories Inside Neural Networks Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention Can SAEs Capture Neural Geometry? Steering Along Manifolds to Control Neural Networks A Geometric Calculator Inside a Neural Network The Neural Geometry Series The World Inside Neural Networks The World Inside Neural Networks Verbalized Eval Awareness Inflates Measured Safety Verbalized Eval Awareness Inflates Measured Safety Paper Summary: Interpreting Language Model Parameters Paper Summary: Interpreting Language Model Parameters Interpreting Language Model Parameters
The Neural Geometry Series
2026-07-08 · via Goodfire Research

The World Inside Neural Networks

The World Inside Neural Networks

How neural geometry will unlock understanding and control of AI

Neural networks develop rich geometric structure in their activations, mirroring the structure of the world they are trained on: days of the week form circles, colors form an HSL manifold, and the tree of life appears in genomic representations. This opening post makes the case that this "neural geometry" is a crucial frontier for understanding, improving, and controlling AI models.

Geiger et al.  ·  May 7, 2026

Steering Along Manifolds to Control Neural Networks

Steering Along Manifolds to Control Neural Networks

Steering along curved manifolds in representation space produces cleaner, more targeted behavior changes than conventional linear steering vectors.

Wurgaft et al.  ·  May 7, 2026

A Geometric Calculator Inside a Neural Network

A Geometric Calculator Inside a Neural Network

We found a neural mechanism that operates over manifolds: a general-purpose addition module inside Llama 3.1 8B which manipulates circular representations of numbers.

Feucht et al.  ·  May 14, 2026

Can SAEs Capture Neural Geometry?

Can SAEs Capture Neural Geometry?

Can we use sparse autoencoder features – i.e., straight lines – to reconstruct curved geometry? We study how, and implement an unsupervised pipeline for discovering manifolds using SAE features.

Bhalla et al.  ·  May 21, 2026

Meandering on Manifolds: The Neural Geometry of Stories Over Time

To fully understand LLM representations, we must understand how they change dynamically over the course of a prompt or conversation. We investigate these temporal dynamics with a simple case study: how do LLMs represent emotions while reading short stories, both geometrically (in activation space) and temporally (changing from sentence to sentence)?

Bigelow et al.  ·  June 23, 2026

Uncovering Neural Geometry in Vision Models With Block-Sparse Featurizers

Uncovering Neural Geometry in Vision Models With Block-Sparse Featurizers

We introduce Block-Sparse Featurizers (BSF), a family of methods to decompose a model's activations into multidimensional subspaces rather than single directions. Applied to vision models, we find that BSFs find interpretable, multidimensional features which offer a more parsimonious explanation of model internals; that those features enable fine-grained steering; and that most concepts in the models are multidimensional.

Fel et al.  ·  July 7, 2026

More posts coming soon!