INDEX

Explanations

patterns that repeat multiple times throughout the document, indicating a specific recurring theme or topic

oai_token-act-pair · gpt-3.5-turbo

terms related to comparisons or similarities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

bang

-0.67

 Derby

-0.64

enders

-0.64

ostic

-0.62

organic

-0.61

ighters

-0.61

ducers

-0.60

usher

-0.60

 Ventura

-0.60

otropic

-0.58

POSITIVE LOGITS

theless

0.90

ctr

0.82

soever

0.79

forward

0.74

 Princ

0.72

tainment

0.71

forth

0.70

isSpecialOrderable

0.69

 situated

0.67

ername

0.67

Activations Density 0.006%

patterns that repeat multiple times throughout the document, indicating a specific recurring theme or topic

terms related to comparisons or similarities

No Comments

No Known Activations