INDEX

Explanations

specific actions like reading, watching, examining, and talking

oai_token-act-pair · gpt-3.5-turbo

actions that involve reading, watching, or reviewing content

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

essert

-0.68

Els

-0.67

equal

-0.65

 harbour

-0.65

amen

-0.63

interstitial

-0.62

arded

-0.62

 unfairly

-0.62

Enjoy

-0.59

IVE

-0.58

POSITIVE LOGITS

hran

0.68

 foregoing

0.67

 assurances

0.67

¿½

0.64

 acquaint

0.62

romy

0.61

reports

0.60

math

0.59

KH

0.59

KS

0.59

Activations Density 0.491%

specific actions like reading, watching, examining, and talking

actions that involve reading, watching, or reviewing content

No Comments

No Known Activations