INDEX

Explanations

hyperlinks or connections between concepts/entities

oai_token-act-pair · gpt-3.5-turbo

mentions of "links" or connections within the text

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

onent

-0.80

creen

-0.66

vation

-0.65

oyer

-0.64

romy

-0.63

pee

-0.62

cry

-0.62

 Apocalypse

-0.61

iva

-0.60

sein

-0.60

POSITIVE LOGITS

 links

3.93

 link

2.71

links

2.69

 Links

2.40

Links

1.98

 ties

1.98

 linking

1.91

 connections

1.90

link

1.82

 LINK

1.76

Activations Density 0.008%

hyperlinks or connections between concepts/entities

mentions of "links" or connections within the text

No Comments

No Known Activations