INDEX

Explanations

terms related to high-level meetings or conferences

oai_token-act-pair · gpt-3.5-turbo

references to summits or high-level meetings

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

enfranch

-0.81

served

-0.76

effic

-0.73

heter

-0.72

relative

-0.71

aver

-0.71

lost

-0.69

lean

-0.68

old

-0.68

uctor

-0.68

POSITIVE LOGITS

 summit

1.46

 Summit

1.00

iership

0.79

 organisers

0.78

 podium

0.74

 volcano

0.73

 Everest

0.72

 sessions

0.72

 boards

0.71

 apex

0.69

Activations Density 0.006%

terms related to high-level meetings or conferences

references to summits or high-level meetings

No Comments

No Known Activations