INDEX

Explanations

names of places or people with the syllable "har" and it gives a high activation value for occurrences of the name "Harri"

oai_token-act-pair · gpt-3.5-turbo

names or terms related to certain places or people

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

 subp

-0.81

İĭ

-0.77

 guarant

-0.64

ulhu

-0.63

ifice

-0.62

 magnification

-0.61

 coord

-0.61

 contrace

-0.60

 carrot

-0.58

 Cosponsors

-0.58

POSITIVE LOGITS

inger

0.79

oola

0.76

ogle

0.73

schild

0.72

raid

0.70

cliffe

0.69

ette

0.66

dal

0.66

ocks

0.66

Collins

0.65

Activations Density 0.092%

names of places or people with the syllable "har" and it gives a high activation value for occurrences of the name "Harri"

names or terms related to certain places or people

No Comments

No Known Activations