INDEX

Explanations

numerical information that involves manipulation or processing of data

oai_token-act-pair · gpt-3.5-turbo

mentions of "input" in various contexts related to data processing

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

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

astern

-0.68

ETA

-0.62

 Peninsula

-0.61

 Seraph

-0.61

ALE

-0.61

neys

-0.61

orks

-0.59

UCK

-0.59

hips

-0.59

APH

-0.59

POSITIVE LOGITS

 input

1.23

 inputs

1.18

input

1.11

ted

0.88

 Input

0.86

ftime

0.86

output

0.84

Input

0.81

 outputs

0.80

ierrez

0.80

Activations Density 0.006%

numerical information that involves manipulation or processing of data

mentions of "input" in various contexts related to data processing

No Comments

No Known Activations