Output Explorer

Every prompt in the paper, and what each model wrote back.

One magic number hidden in filler text; retrieve it.

13 of 100 prompts

Nearby prompts. All 100 Single needle · 8K prompts

PromptRULER niah_single_1 ·8,192 tokens
A special magic number is hidden within the following text. Make sure to memorize it. I will quiz you about the number afterwards.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
[… 109 lines omitted for display …]
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
One of the special magic numbers for wet-lollipop is: 1987737.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
[… 174 lines omitted for display …]
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
The grass is green. The sky is blue. The sun is yellow. Here we go. There and back again.
What is the special magic number for wet-lollipop mentioned in the provided text?
Original prompt 7,658 tokens, 301 linesShown 18 lines around the needleNeedle at 39% depthHaystacks are built per model’s tokenizer, so the needle differs by model
Expected answer
ChronoGPT6543670Aux 20151987737Ours1987737Ours, no steer.1987737PiT-FT6543670Qwen 27B1987737
Models
4 of 4 columns · click a model to add or remove it

Ours

Found the needle

1987737.

13 characters10 tokens

Aux 2015

Found the needle

1987737.

13 characters10 tokens

PiT-FT 2015

Found the needle

: 6543670

9 characters5 tokens

ChronoGPT 2015

Found the needle

: 6543670

9 characters5 tokens