INDEX

Explanations

phrases related to corresponding or related entities

oai_token-act-pair · gpt-3.5-turbo

terms that relate to corresponding relationships and inhabitants of a place

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

 bureaucracy

-0.70

Thu

-0.69

 filibuster

-0.66

 Nationals

-0.63

itism

-0.62

 resume

-0.61

America

-0.60

 hindsight

-0.58

oak

-0.58

odge

-0.57

POSITIVE LOGITS

 corresponding

1.93

 vicinity

1.54

mentioned

1.51

 respective

1.40

 inhabitants

1.38

 aforementioned

1.32

 thereof

1.21

 afore

1.07

 thereto

1.04

 respectively

0.95

Activations Density 0.027%

phrases related to corresponding or related entities

terms that relate to corresponding relationships and inhabitants of a place

No Comments

No Known Activations