INDEX

Explanations

specific numeric values

oai_token-act-pair · gpt-3.5-turbo

occurrences of numerical values and references to fairness

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

xon

-0.72

 shocks

-0.70

itri

-0.68

 Staten

-0.63

houn

-0.63

 bite

-0.62

TN

-0.61

orno

-0.60

olin

-0.60

 Shoals

-0.59

POSITIVE LOGITS

2.50

ï¸ı

1.98

 faire

1.71

1.54

1.38

1.20

tal

1.09

 Talent

1.08

1.06

ï¸

1.05

Activations Density 0.041%

specific numeric values

occurrences of numerical values and references to fairness

No Comments

No Known Activations