INDEX

Explanations

directives or recommendations regarding reading and engagement with content

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 6-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_6.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.6.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

etics

-0.86

actionDate

-0.80

joice

-0.78

merga

-0.76

wind

-0.76

emp

-0.75

emi

-0.74

amar

-0.74

abuse

-0.72

azel

-0.72

POSITIVE LOGITS

 recommendation

0.85

 commodity

0.84

 solution

0.83

 destination

0.83

 scenario

0.82

 performer

0.81

 avenue

0.80

 asset

0.80

 recommendations

0.80

 starter

0.79

Activations Density 0.237%

directives or recommendations regarding reading and engagement with content

No Comments

No Known Activations