INDEX

Explanations

years related to specific events or actions

oai_token-act-pair · gpt-3.5-turbo

references to the year 2016

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

imen

-0.66

angs

-0.63

 clipping

-0.63

FACE

-0.62

ificate

-0.62

Ń·

-0.61

iple

-0.60

 rapport

-0.59

 cumbers

-0.59

stein

-0.59

POSITIVE LOGITS

1.11

bleacher

0.84

âĢ²

0.76

å¹

0.76

 onwards

0.72

âĢ³

0.70

riots

0.67

 Became

0.65

ivari

0.65

 Olympics

0.65

Activations Density 0.067%

years related to specific events or actions

references to the year 2016

No Comments

No Known Activations