INDEX

Explanations

specific references to people meeting or having meetings

oai_token-act-pair · gpt-3.5-turbo

instances of the word "met" in relation to meetings or encounters

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

©¶æ¥µ

-0.89

fighter

-0.75

-0.72

});

-0.70

mington

-0.62

ulo

-0.61

price

-0.60

control

-0.59

margin

-0.59

arily

-0.59

POSITIVE LOGITS

 regards

0.84

 Dull

0.76

 representatives

0.76

 constituents

0.75

 regard

0.75

 strangers

0.75

OTOS

0.74

PsyNetMessage

0.71

draw

0.67

 coworkers

0.67

Activations Density 0.079%

specific references to people meeting or having meetings

instances of the word "met" in relation to meetings or encounters

No Comments

No Known Activations