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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

itud

-0.69

eddy

-0.69

 Qiao

-0.66

uably

-0.65

esty

-0.64

lb

-0.63

itudes

-0.63

ÙĴ

-0.63

 doubling

-0.62

Whe

-0.62

POSITIVE LOGITS

 casinos

0.77

 acron

0.72

 unsus

0.71

 sake

0.68

rises

0.66

 tranqu

0.66

 secrecy

0.64

anqu

0.64

 arrivals

0.64

 patronage

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.