INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 departure

-0.73

 unch

-0.69

 extrad

-0.69

bidden

-0.69

 paths

-0.68

 selves

-0.68

 anonymity

-0.68

ay

-0.68

 departures

-0.67

vir

-0.66

POSITIVE LOGITS

 Cosponsors

0.79

LED

0.73

Tib

0.70

zig

0.70

ellect

0.69

 Improvement

0.69

ighed

0.68

ificantly

0.68

ront

0.68

 Repair

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.