INDEX

Explanations

ranges or numerical values

oai_token-act-pair · gpt-3.5-turbo

numerical ranges or percentages

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

authorized

-0.61

estate

-0.60

wordpress

-0.60

 erected

-0.59

ilyn

-0.59

pain

-0.58

 Polk

-0.57

 PUBLIC

-0.56

conserv

-0.56

 Signs

-0.54

POSITIVE LOGITS

ousand

0.79

mph

0.73

PF

0.72

ãĤ´ãĥ³

0.72

0.71

HK

0.70

tenance

0.70

 sidx

0.69

AM

0.69

GHz

0.69

Activations Density 0.089%

ranges or numerical values

numerical ranges or percentages

No Comments

No Known Activations