INDEX

Explanations

timestamps or instructions related to specific actions or events

oai_token-act-pair · gpt-3.5-turbo

the word "when" indicating specific conditions or time-related phrases

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

oslov

-0.74

eur

-0.73

earances

-0.73

amental

-0.73

whatever

-0.72

ulture

-0.70

zar

-0.69

orate

-0.69

uably

-0.69

zek

-0.68

POSITIVE LOGITS

 encountering

1.28

 comparing

1.23

 dealing

1.21

 attempting

1.20

 applying

1.19

 interacting

1.18

 entering

1.16

 selecting

1.14

 calculating

1.14

 evaluating

1.13

Activations Density 0.088%

timestamps or instructions related to specific actions or events

the word "when" indicating specific conditions or time-related phrases

No Comments

No Known Activations