INDEX

Explanations

words related to medical conditions or terms

oai_token-act-pair · gpt-3.5-turbo

sufficiently frequent occurrences of the suffix "ic."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ĸļ

-0.77

 Parenthood

-0.77

 Kinnikuman

-0.73

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

-0.71

 Sorceress

-0.70

å§«

-0.70

ABE

-0.69

é¾

-0.68

 Colt

-0.67

sold

-0.67

POSITIVE LOGITS

otine

1.02

adic

0.92

entric

0.92

otin

0.90

idal

0.90

otic

0.84

ornia

0.84

atically

0.83

 acid

0.81

illin

0.81

Activations Density 0.030%

words related to medical conditions or terms

sufficiently frequent occurrences of the suffix "ic."

No Comments

No Known Activations