INDEX

Explanations

words related to tracking and monitoring activities or objects

oai_token-act-pair · gpt-3.5-turbo

references to "track" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

fter

-0.67

eus

-0.67

KO

-0.65

 Cousins

-0.65

Hua

-0.64

fters

-0.63

cu

-0.63

lations

-0.62

Beg

-0.61

ervative

-0.61

POSITIVE LOGITS

 track

4.02

 Track

2.96

track

2.77

 tracks

2.70

Track

2.59

 Tracks

1.97

 tracked

1.87

tracks

1.80

 tracker

1.63

 tracking

1.54

Activations Density 0.018%

words related to tracking and monitoring activities or objects

references to "track" in various contexts

No Comments

No Known Activations