INDEX

Explanations

numerical information structured in a specific way, often involving rankings or statistics

oai_token-act-pair · gpt-3.5-turbo

references to rankings or standings in competitive contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ahime

-0.74

ymm

-0.69

rance

-0.69

ouf

-0.63

 Fenrir

-0.61

-0.60

usher

-0.60

odium

-0.60

cker

-0.58

 Qiao

-0.58

POSITIVE LOGITS

 past

0.96

 last

0.80

osures

0.79

Last

0.72

top

0.71

 remaining

0.71

 next

0.70

teenth

0.69

 nineteen

0.69

Activations Density 0.073%

numerical information structured in a specific way, often involving rankings or statistics

references to rankings or standings in competitive contexts

No Comments

No Known Activations