INDEX

Explanations

informational phrases for various topics or organizations

oai_token-act-pair · gpt-3.5-turbo

phrases that provide information or references to additional information sources

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

aud

-0.71

meter

-0.70

xtap

-0.62

alone

-0.61

Sort

-0.61

 choir

-0.61

adj

-0.60

 Dull

-0.60

 Gong

-0.58

 Bastard

-0.57

POSITIVE LOGITS

 regarding

0.84

 details

0.77

 info

0.76

tackle

0.74

 pertaining

0.73

---------

0.71

 concerning

0.70

finder

0.70

utory

0.69

 relating

0.69

Activations Density 0.035%

informational phrases for various topics or organizations

phrases that provide information or references to additional information sources

No Comments

No Known Activations