INDEX

Explanations

instances where something is done accurately or precisely

oai_token-act-pair · gpt-3.5-turbo

instances of the word "correctly" and emphasize issues of correctness and accuracy

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

oleon

-0.75

RAW

-0.67

hyde

-0.67

ergus

-0.64

 Alam

-0.62

iments

-0.62

INGS

-0.60

 Valhalla

-0.59

ires

-0.59

 Skydragon

-0.58

POSITIVE LOGITS

 correctly

1.08

correct

1.06

 corrected

0.97

 correct

0.91

Correct

0.90

 regenerate

0.84

 aligned

0.82

 alloc

0.82

 propagate

0.80

 behaved

0.77

Activations Density 0.004%

instances where something is done accurately or precisely

instances of the word "correctly" and emphasize issues of correctness and accuracy

No Comments

No Known Activations