INDEX

Explanations

affirmative and negative indicators, often associated with binary choices or confirmations

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

Not in Any Lists

No Comments

Negative Logits

ongyang

-0.73

vier

-0.71

awei

-0.70

soever

-0.69

orse

-0.67

advant

-0.66

erred

-0.66

xual

-0.65

resses

-0.64

oward

-0.64

POSITIVE LOGITS

 è£ıè

1.04

ahime

0.67

});

0.67

Cub

0.62

 Minotaur

0.62

ACP

0.61

 Kafka

0.61

ä¹

0.61

aez

0.59

ãĤ¨ãĥ«

0.59

Activations Density 0.089%

affirmative and negative indicators, often associated with binary choices or confirmations

No Comments

No Known Activations