INDEX

Explanations

statements mentioning actions or events as reported or believed to have occurred

oai_token-act-pair · gpt-3.5-turbo

phrases indicating possession or existence

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

fortunately

-0.74

ppings

-0.73

 Timbers

-0.71

atters

-0.70

Reviewer

-0.67

ortunately

-0.66

 Bucs

-0.65

 Bundy

-0.64

clip

-0.63

 Tacoma

-0.63

POSITIVE LOGITS

leeve

0.73

 categor

0.73

 âĢİ

0.72

 influenced

0.68

 accounted

0.67

igious

0.67

 exempt

0.67

 operated

0.65

 composed

0.65

 driven

0.64

Activations Density 0.075%

statements mentioning actions or events as reported or believed to have occurred

phrases indicating possession or existence

No Comments

No Known Activations