INDEX

Explanations

numbers related to rankings

oai_token-act-pair · gpt-3.5-turbo

references to rankings or numerical positions

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

iership

-0.79

RAFT

-0.68

assies

-0.68

rez

-0.67

ãĤ¼ãĤ¦ãĤ¹

-0.67

romeda

-0.66

rosse

-0.65

aven

-0.64

wered

-0.64

inese

-0.64

POSITIVE LOGITS

xious

1.00

except

0.87

obs

0.85

AH

0.84

ERROR

0.81

zzle

0.79

 longer

0.79

 Matter

0.77

 matter

0.77

 doubt

0.74

Activations Density 0.028%

numbers related to rankings

references to rankings or numerical positions

No Comments

No Known Activations