Seeking arXiv cs.CL Endorsement for My Research Paper (SRRM: A Retrieval-Reconstruction Benchmark for Evaluating Long-Context Memory in Small Language Models)

Research
1

Hello everyone,

I am preparing to submit my first research paper to arXiv (cs.CL) and am currently looking for an eligible endorser.

The paper introduces Simple Retrieval-Reconstruction Memory (SRRM), a lightweight and fully reproducible benchmark for evaluating long-context memory in Small Language Models. Unlike existing retrieval-oriented benchmarks, SRRM jointly evaluates:

  • Targeted Retrieval
  • Global Information Reconstruction
  • Memory Preservation Ratio (MPR), a benchmark-level metric relating retrieval and reconstruction performance.

The benchmark was evaluated on the Qwen2.5 model family (0.5B, 1.5B, and 3B) under progressively increasing context loads (10–100 factual statements). Interestingly, the benchmark consistently showed that the Qwen2.5-1.5B model outperformed the larger 3B model across retrieval accuracy, reconstruction performance, and MPR, suggesting that parameter count alone does not necessarily improve long-context memory preservation.

If you have experience publishing under cs.CL, I would greatly appreciate your feedback on the manuscript. If, after reviewing the paper, you believe it is appropriate for the category and are eligible to provide an arXiv endorsement, I would sincerely appreciate your consideration.

Thank you for your time.