Hugging Face's logo

Datasets:
giantfish-fly
/
pi-llm

Tasks:
Question Answering
Modalities:
Tabular
Text
Languages:
English
Size:
< 1K
ArXiv:
Tags:
llm
memory
retrieval
context interference
long-context
License:
Dataset card Data Studio Files
xet
Community
1
giantfish-fly commited on
Commit
54fbb51
·
verified ·
1 Parent(s): d55e3e3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -3
README.md CHANGED
@@ -31,11 +31,9 @@ configs:
31
  # PI-LLM Bench: The Core Retrieval Challenge Behind MRCR
32
  - ICML 2025 Long-Context Foundation Models Workshop Accepted.
33
 
34
- >A simple context interference evaluation.
35
  > **Adoption (Aug 31, 2025):** Officially integrated into a top-5 open-weight model company’s **internal benchmarking framework** for assessing ** tracking capacity and context interference in agents**.
36
 
37
-
38
-
39
  ## TL;DR
40
  We identify a task that is **super easy for humans** but where all LLMs—from early 0.1B to the most modern 600B+ (GPT-5, Grok-4, Gemini, DeepSeek, etc.)—consistently **fail in the Same Way**. This pinpoints the **core challenge of MRCR**
41
 
 
31
  # PI-LLM Bench: The Core Retrieval Challenge Behind MRCR
32
  - ICML 2025 Long-Context Foundation Models Workshop Accepted.
33
 
34
+ A simple context interference evaluation.
35
  > **Adoption (Aug 31, 2025):** Officially integrated into a top-5 open-weight model company’s **internal benchmarking framework** for assessing ** tracking capacity and context interference in agents**.
36
 
 
 
37
  ## TL;DR
38
  We identify a task that is **super easy for humans** but where all LLMs—from early 0.1B to the most modern 600B+ (GPT-5, Grok-4, Gemini, DeepSeek, etc.)—consistently **fail in the Same Way**. This pinpoints the **core challenge of MRCR**
39