INDEX

Explanations

the word "liked" followed by various subjects or objects

oai_token-act-pair · gpt-3.5-turbo

mentions of preferences or feelings towards something, primarily focusing on the word "liked."

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Terminal

-0.63

 extingu

-0.62

 stage

-0.61

 recovery

-0.60

 Elim

-0.58

 eviction

-0.57

 mounting

-0.56

 Express

-0.56

olve

-0.55

 sovereignty

-0.54

POSITIVE LOGITS

 liked

3.37

 disliked

2.11

 loved

1.95

 enjoyed

1.71

 hated

1.71

 likes

1.67

 liking

1.64

 admired

1.62

 appreciated

1.44

 Likes

1.36

Activations Density 0.012%

the word "liked" followed by various subjects or objects

mentions of preferences or feelings towards something, primarily focusing on the word "liked."

No Comments

No Known Activations