INDEX

Explanations

phrases related to complete entities or processes

oai_token-act-pair · gpt-3.5-turbo

components related to overarching themes or concepts in various contexts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 occasional

-0.73

cest

-0.73

uly

-0.67

 slightest

-0.66

earcher

-0.65

 Friendly

-0.65

egu

-0.61

 Fighters

-0.60

 cents

-0.60

 Friend

-0.59

POSITIVE LOGITS

 except

0.91

 revolves

0.90

 Including

0.83

ä»£

0.77

 including

0.75

âĶĢâĶĢ

0.74

 saga

0.72

insula

0.72

 catalogue

0.71

 hinges

0.71

Activations Density 0.210%

phrases related to complete entities or processes

components related to overarching themes or concepts in various contexts

No Comments

No Known Activations