INDEX

Explanations

the word "ton" with high activation values

oai_token-act-pair · gpt-3.5-turbo

the repeated mention of the word "ton."

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

 Phant

-0.71

Craft

-0.70

 Danger

-0.67

 Cust

-0.67

ELD

-0.67

Constructed

-0.67

 Clinic

-0.63

Journal

-0.61

 Forever

-0.61

pring

-0.60

POSITIVE LOGITS

neau

1.05

ne

0.92

arg

0.90

Ton

0.89

aton

0.89

ights

0.87

umen

0.87

eful

0.86

 earthqu

0.86

icum

0.85

Activations Density 0.018%

the word "ton" with high activation values

the repeated mention of the word "ton."

No Comments

No Known Activations