INDEX

Explanations

technical, possibly medical, terminology in a structured format

oai_token-act-pair · gpt-3.5-turbo

repeated phrases or patterns indicating a consistent sentiment or statement

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

 Kenyan

-0.76

 funer

-0.71

 scattering

-0.64

 Tanz

-0.64

 Sicily

-0.64

 Libyan

-0.63

 cellphone

-0.63

 reception

-0.62

 vomiting

-0.62

 Tunis

-0.62

POSITIVE LOGITS

º

0.89

Â¯

0.81

dro

0.81

Pg

0.80

should

0.79

agree

0.79

erest

0.79

âģ

0.77

į

0.76

hur

0.75

Activations Density 0.175%

technical, possibly medical, terminology in a structured format

repeated phrases or patterns indicating a consistent sentiment or statement

No Comments

No Known Activations