INDEX

Explanations

dates and specific time references

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.01

1:0.01

2:0.04

3:0.05

4:0.05

5:0.02

6:0.10

7:0.41

8:0.03

9:0.04

10:0.10

11:0.07

Negative Logits

vernment

-1.97

��極

-1.94

版

-1.74

taboola

-1.72

arrow

-1.69

english

-1.69

asive

-1.62

cpp

-1.59

arers

-1.57

 widen

-1.57

POSITIVE LOGITS

 birthday

1.70

 Thanksgiving

1.64

 dinners

1.58

 Sabbath

1.54

 Feast

1.54

 dinner

1.48

 Remem

1.47

 sunrise

1.46

 brunch

1.45

Mut

1.45

Activations Density 0.014%

dates and specific time references

No Known Activations