INDEX

Explanations

terms related to verbal communication and abuse

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

encies

-0.80

theme

-0.73

nesota

-0.73

âĨĳ

-0.69

bard

-0.69

Sec

-0.67

sites

-0.65

 Emin

-0.64

 Anthem

-0.64

addafi

-0.63

POSITIVE LOGITS

 manifestation

0.72

urn

0.70

\\\\\\\\

0.69

reet

0.67

 copies

0.66

 embodiment

0.65

 proofs

0.63

SN

0.63

iple

0.63

ĵ

0.62

Activations Density 0.113%

terms related to verbal communication and abuse

No Comments

No Known Activations