INDEX

Explanations

phrases indicating future actions or plans

oai_token-act-pair · gpt-3.5-turbo

instances of the word "be" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

emonic

-0.65

 Bars

-0.64

 promise

-0.63

ossus

-0.63

might

-0.63

usc

-0.62

 Oops

-0.62

 theorem

-0.61

arta

-0.61

POSITIVE LOGITS

 able

1.05

 replaced

1.00

ige

0.97

 rewarded

0.96

heading

0.94

 remembered

0.92

 phased

0.90

fitting

0.90

 judged

0.90

AMS

0.88

Activations Density 0.162%

phrases indicating future actions or plans

instances of the word "be" in various contexts

No Comments

No Known Activations