INDEX

Explanations

references to the word "shark"

oai_token-act-pair · gpt-3.5-turbo

mentions of sharks

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

 Dayton

-0.78

orter

-0.74

cu

-0.72

Whe

-0.67

 Eisenhower

-0.66

 Ohio

-0.65

 Davis

-0.64

COL

-0.64

Cel

-0.64

 Advent

-0.64

POSITIVE LOGITS

 shark

3.74

 sharks

3.52

 Shark

3.29

 Sharks

2.60

 whale

1.76

 whales

1.64

 Whale

1.60

 fins

1.59

 squid

1.52

 shrimp

1.45

Activations Density 0.019%

references to the word "shark"

mentions of sharks

No Comments

No Known Activations