INDEX

Explanations

names related to different individuals or characters

oai_token-act-pair · gpt-3.5-turbo

references to specific names and terms associated with individuals and events

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

fman

-0.87

eln

-0.80

 gunned

-0.77

quartered

-0.74

enture

-0.70

ounced

-0.69

ele

-0.69

icably

-0.68

 transpl

-0.68

ed

-0.67

POSITIVE LOGITS

Magikarp

0.80

 Lank

0.78

 Flavoring

0.77

osity

0.74

adoes

0.72

ãĥ³ãĤ¸

0.70

itor

0.69

erville

0.69

ãĤ¤ãĥĪ

0.67

ï¸

0.67

Activations Density 0.042%

names related to different individuals or characters

references to specific names and terms associated with individuals and events

No Comments

No Known Activations