INDEX

Explanations

proper nouns or adjectives related to people or events being admired or honored

oai_token-act-pair · gpt-3.5-turbo

instances of the word "celebrated" or related forms indicating recognition or commemoration

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

adra

-1.00

heed

-0.71

ings

-0.70

clerosis

-0.68

emen

-0.68

DoS

-0.67

thora

-0.66

eper

-0.66

inki

-0.65

adas

-0.65

POSITIVE LOGITS

 Seym

0.90

CLASSIFIED

0.76

nesday

0.74

 externalToEVAOnly

0.74

theless

0.71

 BaseType

0.70

hester

0.69

 Pulitzer

0.67

 trumpet

0.65

urated

0.64

Activations Density 0.028%

proper nouns or adjectives related to people or events being admired or honored

instances of the word "celebrated" or related forms indicating recognition or commemoration

No Comments

No Known Activations