INDEX

Explanations

instances of names and identifiers associated with specific entities or organizations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.10.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

omics

-0.76

ecd

-0.68

issions

-0.68

advertising

-0.66

estyles

-0.65

raged

-0.64

eties

-0.63

lust

-0.63

tails

-0.63

rules

-0.63

POSITIVE LOGITS

 culprit

0.98

 protagonist

0.81

 inhabit

0.81

 entity

0.81

 deity

0.80

 genus

0.80

 inventor

0.77

 antagonist

0.76

 perpetrator

0.75

 person

0.74

Activations Density 0.988%

instances of names and identifiers associated with specific entities or organizations

No Comments

No Known Activations