INDEX

Explanations

words related to expressing conclusions or introducing explanations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "So" indicating a transition or conclusion in thought

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

 repr

-0.66

 arch

-0.64

 appearing

-0.64

 marks

-0.63

 following

-0.63

 unknown

-0.62

 mark

-0.61

 borrowed

-0.61

 display

-0.60

 deton

-0.60

POSITIVE LOGITS

So

3.02

Anyway

2.09

Therefore

1.96

So

1.92

Thus

1.92

Now

1.67

And

1.65

so

1.62

Okay

1.60

But

1.56

Activations Density 0.018%

words related to expressing conclusions or introducing explanations

instances of the word "So" indicating a transition or conclusion in thought

No Comments

No Known Activations