INDEX

Explanations

phrases indicating written communication, specifically letters and requests

oai_token-act-pair · gpt-3.5-turbo

instances of correspondence or communication directed towards specific entities or individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

ahime

-0.71

tions

-0.66

 bets

-0.65

 calculus

-0.64

ibles

-0.64

 ILCS

-0.62

lux

-0.61

inventoryQuantity

-0.61

nels

-0.61

 crawled

-0.61

POSITIVE LOGITS

 accompany

1.05

 superiors

0.95

 subscribers

0.86

 shareholders

0.85

 congratulate

0.82

 commemorate

0.81

 bombard

0.78

 solicit

0.77

 complain

0.75

 inbox

0.75

Activations Density 0.095%

phrases indicating written communication, specifically letters and requests

instances of correspondence or communication directed towards specific entities or individuals

No Comments

No Known Activations