INDEX

Explanations

dates in the format YYYY appearing in the text

oai_token-act-pair · gpt-3.5-turbo

references to the year 2011

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

phy

-0.84

onite

-0.82

ntil

-0.80

inates

-0.79

imore

-0.75

phia

-0.75

oute

-0.74

odies

-0.72

 cumbers

-0.71

olean

-0.71

POSITIVE LOGITS

å¹

1.02

 onwards

0.80

-'

0.79

ãĥ¼ãĥĨãĤ£

0.78

0.71

worthiness

0.71

 Petraeus

0.69

0.68

ford

0.67

aldo

0.67

Activations Density 0.023%

dates in the format YYYY appearing in the text

references to the year 2011

No Comments

No Known Activations