INDEX

Explanations

numeric fractions in the form of 'two-thirds'

oai_token-act-pair · gpt-3.5-turbo

phrases or data points that reference proportions, specifically relating to two-thirds of a whole

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

tag

-0.64

bru

-0.63

 talk

-0.62

 model

-0.62

 romantic

-0.61

br

-0.61

 pedd

-0.61

 exchange

-0.60

BR

-0.58

 breed

-0.58

POSITIVE LOGITS

thirds

4.04

 thirds

2.42

fif

1.86

quarters

1.69

third

1.53

fifth

1.18

fourth

1.13

half

1.10

majority

1.09

dozen

1.05

Activations Density 0.014%

numeric fractions in the form of 'two-thirds'

phrases or data points that reference proportions, specifically relating to two-thirds of a whole

No Comments

No Known Activations