INDEX

Explanations

from the activations shown, it seems like this neuron is looking for words ending with "-um"

oai_token-act-pair · gpt-3.5-turbo

occurrences of the token "um."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 cutoff

-0.72

 strawberries

-0.69

âĸĪâĸĪâĸĪâĸĪâĸĪâĸĪâĸĪâĸĪ

-0.67

 Aires

-0.65

 elves

-0.65

Ń·

-0.64

 Shades

-0.63

 blackout

-0.62

jri

-0.62

 Morales

-0.62

POSITIVE LOGITS

osity

1.09

mers

1.09

ming

1.06

essage

0.98

etric

0.98

atism

0.97

mit

0.97

um

0.97

pty

0.96

antic

0.96

Activations Density 0.015%

from the activations shown, it seems like this neuron is looking for words ending with "-um"

occurrences of the token "um."

No Comments

No Known Activations