INDEX

Explanations

the name "Ker" with varying degrees of emphasis or importance

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of the name "Ker."

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

UME

-0.83

ablishment

-0.80

Uncommon

-0.72

BOOK

-0.72

ISION

-0.71

IBLE

-0.71

ï¿½ï¿½ï¿½ï¿½

-0.70

UCT

-0.68

ÙĲ

-0.68

IRD

-0.67

POSITIVE LOGITS

rigan

1.14

sten

1.05

bil

0.96

oshenko

0.93

Ker

0.91

lins

0.88

 horizont

0.87

grim

0.85

©¶æ

0.84

mit

0.84

Activations Density 0.005%

the name "Ker" with varying degrees of emphasis or importance

repeated mentions of the name "Ker."

No Comments

No Known Activations