INDEX

Explanations

medical terms and procedures related to health care

oai_token-act-pair · gpt-3.5-turbo

conjunctions that indicate lists or connections between ideas

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Saying

-0.67

js

-0.67

-0.64

rb

-0.62

aldo

-0.62

lear

-0.62

hart

-0.62

jar

-0.61

Redditor

-0.60

heads

-0.60

POSITIVE LOGITS

 assorted

1.15

 possibly

0.90

 downright

0.87

 ultimately

0.86

 even

0.85

 finally

0.83

 other

0.80

whatever

0.78

 others

0.76

cellaneous

0.74

Activations Density 0.153%

medical terms and procedures related to health care

conjunctions that indicate lists or connections between ideas

No Comments

No Known Activations