INDEX

Explanations

Hmm, it appears that Neuron 4 is looking for patterns from a data source that may be encoded in a different format, possibly related to a specific language or system, as there doesn't seem to be a clear pattern based on the text provided

oai_token-act-pair · gpt-3.5-turbo

special characters or symbols with specific patterns

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

theless

-0.71

aturdays

-0.70

ographically

-0.67

ulators

-0.67

 Kens

-0.66

ithing

-0.64

 Ness

-0.63

 Redd

-0.63

anwhile

-0.61

essa

-0.60

POSITIVE LOGITS

¹

1.95

²

1.88

¨

1.86

°

1.85

Ń

1.84

¶

1.83

¦

1.79

´

1.76

¸

1.76

¯

1.76

Activations Density 0.007%

Hmm, it appears that Neuron 4 is looking for patterns from a data source that may be encoded in a different format, possibly related to a specific language or system, as there doesn't seem to be a clear pattern based on the text provided

special characters or symbols with specific patterns

No Comments

No Known Activations