INDEX

Explanations

mentions of specific political figures, particularly related to government actions and statements

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "House" or "Senate" indicating political context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 checkpoint

-0.72

 Darwin

-0.70

 radial

-0.67

 value

-0.67

 lateral

-0.66

 evaluation

-0.65

 athe

-0.64

 distances

-0.64

tal

-0.63

 doping

-0.63

POSITIVE LOGITS

House

3.74

house

2.06

Senate

2.00

 HOUSE

2.00

 House

1.79

houses

1.65

HOU

1.59

 Houses

1.45

Republicans

1.42

Room

1.35

Activations Density 0.014%

mentions of specific political figures, particularly related to government actions and statements

occurrences of the word "House" or "Senate" indicating political context

No Comments

No Known Activations