INDEX

Explanations

instances where something is replaced or done differently than expected

oai_token-act-pair · gpt-3.5-turbo

phrases that signal a shift or contrast in circumstances or actions

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

 lapt

-0.62

breakers

-0.62

lees

-0.61

cycl

-0.61

 neighbourhood

-0.60

chet

-0.60

 Quake

-0.59

emate

-0.59

 Sahara

-0.58

wel

-0.58

POSITIVE LOGITS

Ļ½

0.76

ocus

0.75

 chose

0.75

 opting

0.72

 instead

0.72

opt

0.71

nels

0.71

 opted

0.71

 relied

0.67

chart

0.66

Activations Density 0.019%

instances where something is replaced or done differently than expected

phrases that signal a shift or contrast in circumstances or actions

No Comments

No Known Activations