INDEX

Explanations

terms related to weather phenomena, particularly fog

oai_token-act-pair · gpt-3.5-turbo

mentions of fog and related imagery

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

aza

-0.75

 wrists

-0.75

PB

-0.74

 Egyptians

-0.73

 Mobility

-0.73

 Spart

-0.73

anyl

-0.70

reads

-0.69

 Nasa

-0.69

odus

-0.69

POSITIVE LOGITS

fog

2.14

Fog

1.86

Rag

1.27

Fay

1.21

 distortion

1.18

 Lambert

1.18

 fuzz

1.14

 muff

1.11

 clut

1.10

hed

1.01

Activations Density 0.058%

terms related to weather phenomena, particularly fog

mentions of fog and related imagery

No Comments

No Known Activations