INDEX

Explanations

capitalized terms like 'REWARD', 'RECOMMENDED', and 'REVIEW'

oai_token-act-pair · gpt-3.5-turbo

instances of the abbreviation "RE" indicating reviews or references

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

glers

-0.79

Ic

-0.74

 Indonesian

-0.74

ãĥīãĥ©ãĤ´ãĥ³

-0.73

ãĤ¨ãĥ«

-0.70

 Icelandic

-0.70

 Konami

-0.69

 Kissinger

-0.69

 Finnish

-0.68

 Kafka

-0.67

POSITIVE LOGITS

LEASE

1.21

ACH

1.13

VERS

1.12

PRES

1.10

VOL

1.09

WARD

1.09

VIEW

1.08

QUEST

1.07

FER

1.05

ASON

1.04

Activations Density 0.008%

capitalized terms like 'REWARD', 'RECOMMENDED', and 'REVIEW'

instances of the abbreviation "RE" indicating reviews or references

No Comments

No Known Activations