INDEX

Explanations

profane words or phrases

oai_token-act-pair · gpt-3.5-turbo

linguistic patterns related to specific suffixes and potentially notable keywords

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

rift

-0.80

iry

-0.75

 Wond

-0.74

MQ

-0.74

romy

-0.70

dq

-0.69

DOM

-0.67

IR

-0.67

Soc

-0.66

ilk

-0.64

POSITIVE LOGITS

 banter

0.82

 insult

0.78

 gluc

0.74

 fres

0.74

plet

0.74

 heck

0.73

eah

0.72

boo

0.66

onsense

0.66

 outburst

0.66

Activations Density 0.055%

profane words or phrases

linguistic patterns related to specific suffixes and potentially notable keywords

No Comments

No Known Activations