INDEX

Explanations

references to actions or conditions described as "above" with numerical values associated

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

Pwr

-0.75

emis

-0.72

atown

-0.72

eday

-0.71

mates

-0.70

 Speedway

-0.70

paio

-0.69

ĪĴ

-0.68

ULTS

-0.67

ĳå£«

-0.66

POSITIVE LOGITS

mentioned

1.46

 scenario

0.92

 mentioned

0.89

 phenomenon

0.84

 phenomena

0.83

 quote

0.81

 quotation

0.81

 statement

0.80

 timeframe

0.78

 aforementioned

0.78

Activations Density 0.060%

references to actions or conditions described as "above" with numerical values associated

No Known Activations