INDEX

Explanations

terms related to specific characters or concepts that appear to be in a different language

oai_token-act-pair · gpt-3.5-turbo

special characters or symbols that may indicate specific formatting or emphasis

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 chunks

-0.69

 loopholes

-0.69

 expectancy

-0.67

 patched

-0.67

 unborn

-0.66

 eyeb

-0.66

 tuna

-0.66

 Turing

-0.65

 chained

-0.64

 censored

-0.64

POSITIVE LOGITS

ï¸ı

1.16

ng

0.98

ti

0.98

ski

0.97

Å«

0.97

eh

0.97

¡

0.95

Â§

0.95

ller

0.95

Å

0.95

Activations Density 0.055%

terms related to specific characters or concepts that appear to be in a different language

special characters or symbols that may indicate specific formatting or emphasis

No Comments

No Known Activations