INDEX

Explanations

violent actions like sabotage and infiltration

oai_token-act-pair · gpt-3.5-turbo

terms related to sabotage and infiltration

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

eral

-0.73

cil

-0.72

area

-0.72

Lear

-0.71

idd

-0.70

cester

-0.70

pool

-0.70

âĢ¢âĢ¢âĢ¢âĢ¢

-0.70

Ice

-0.70

Foot

-0.69

POSITIVE LOGITS

 sabotage

1.06

 sabot

0.90

 infiltr

0.78

 infiltrate

0.72

 torpedo

0.72

 tactics

0.70

abwe

0.69

 Rebel

0.68

xual

0.66

 infiltration

0.65

Activations Density 0.034%

violent actions like sabotage and infiltration

terms related to sabotage and infiltration

No Comments

No Known Activations