INDEX

Explanations

occurrences of phrases indicating events, experiences, or features related to entertainment and media

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

ignore

-0.68

ĺħ

-0.67

 equilibrium

-0.64

 objections

-0.64

pillar

-0.64

 exceptions

-0.64

 subcontract

-0.64

 notwithstanding

-0.62

Redditor

-0.62

enough

-0.61

POSITIVE LOGITS

 unforgettable

0.91

 thrilling

0.89

 exciting

0.88

 FREE

0.88

 exhilar

0.87

 curated

0.83

eaturing

0.81

 beautifully

0.80

 exclusive

0.80

 unparalleled

0.79

Activations Density 0.824%

occurrences of phrases indicating events, experiences, or features related to entertainment and media

No Comments

No Known Activations