INDEX

Explanations

years

oai_token-act-pair · gpt-3.5-turbo

references to specific years and statistical achievements in sports contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

atom

-0.72

abulary

-0.68

 shutter

-0.66

ansk

-0.61

ndra

-0.59

 ancest

-0.58

avis

-0.58

izen

-0.58

zsche

-0.58

eping

-0.58

POSITIVE LOGITS

 playoffs

1.25

 Rookie

1.18

-'

1.14

 postseason

1.07

 Playoffs

1.04

 playoff

1.04

 season

0.98

 STATS

0.91

 rookie

0.90

 championship

0.89

Activations Density 0.077%

years

references to specific years and statistical achievements in sports contexts

No Comments

No Known Activations