INDEX

Explanations

sources or attributions at the end of documents or statements

oai_token-act-pair · gpt-3.5-turbo

references to sources of information or citations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

rift

-0.69

lights

-0.67

 wrench

-0.66

 stand

-0.65

 balancing

-0.63

 sharp

-0.63

 itch

-0.61

 blot

-0.60

 marks

-0.60

 rise

-0.60

POSITIVE LOGITS

Via

4.22

Via

2.31

via

1.91

SOURCE

1.32

kefeller

1.06

Sources

1.05

via

1.03

Shares

1.00

Redditor

1.00

Images

0.96

Activations Density 0.015%

sources or attributions at the end of documents or statements

references to sources of information or citations

No Comments

No Known Activations