INDEX

Explanations

proper nouns, particularly names associated with individuals

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.8.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Negative Logits

tons

-0.73

stown

-0.67

 Amon

-0.64

abad

-0.62

issance

-0.60

 Despair

-0.59

 orbs

-0.59

åŃĲ

-0.57

aceae

-0.57

;;;;;;;;;;;;

-0.56

POSITIVE LOGITS

arrison

0.61

urtle

0.60

ipper

0.60

CEO

0.57

Loading

0.56

<+

0.56

pill

0.56

olester

0.54

ayer

0.54

ECD

0.53

Activations Density 0.414%

proper nouns, particularly names associated with individuals

No Known Activations