INDEX

Explanations

dates or months mentioned within texts

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of "last" and "we" in various contexts

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

text markers or tokens signaling the beginning of a news or reporting context, like "Last", "An", "We", "What".

oai_token-act-pair · claude-3-5-haiku-20241022 Triggered by @cedarren

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 thereafter

-0.70

 thereof

-0.65

 respons

-0.62

pled

-0.62

 versa

-0.61

_.

-0.60

 outwe

-0.59

 comprom

-0.59

 thereto

-0.59

 afterward

-0.59

POSITIVE LOGITS

 Expand

0.85

 Introduction

0.79

 Vegan

0.77

 SHARES

0.77

 Summary

0.76

 Calculator

0.75

 Transcript

0.74

 Answer

0.73

 Definition

0.72

 Finder

0.71

Activations Density 0.347%

dates or months mentioned within texts

repeated mentions of "last" and "we" in various contexts

text markers or tokens signaling the beginning of a news or reporting context, like "Last", "An", "We", "What".

No Comments

No Known Activations