INDEX

Explanations

actions or steps to accomplish a task

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate sequential actions or instructions

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

 Lauder

-0.74

 Mong

-0.74

TextColor

-0.74

 Taiwanese

-0.69

 millenn

-0.67

 Nanto

-0.67

Thu

-0.65

always

-0.64

 Surveillance

-0.64

////////////////

-0.63

POSITIVE LOGITS

hran

0.86

 mastered

0.77

orest

0.77

 arrives

0.76

acqu

0.76

ividual

0.73

sett

0.73

 arrive

0.73

 foothold

0.72

bered

0.72

Activations Density 0.112%

actions or steps to accomplish a task

phrases that indicate sequential actions or instructions

No Comments

No Known Activations