INDEX

Explanations

references to numbers or figures, particularly repetitions of the number 800 at the highest activation level

oai_token-act-pair · gpt-3.5-turbo

instances of a specific numerical sequence, particularly variations of "800."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

lict

-0.81

uked

-0.77

bid

-0.76

rely

-0.72

 isot

-0.69

mast

-0.68

lying

-0.68

ributed

-0.67

 handc

-0.66

 shepherd

-0.66

POSITIVE LOGITS

1.08

0.95

0.86

MHz

0.85

0.84

0.83

æ©Ł

0.82

 msec

0.81

0.80

Activations Density 0.012%

references to numbers or figures, particularly repetitions of the number 800 at the highest activation level

instances of a specific numerical sequence, particularly variations of "800."

No Comments

No Known Activations