The authors compared two versions of multi-stage pipelines on GSM-Hard and MATH-500.
LLM pipelines can lose 40.5 accuracy points between stages
Across mathematical reasoning benchmarks, preserving access to the original problem substantially improved multi-stage pipelines without changing the model, prompts, stages or completion budget.
Top University
Tianqi Bu · YuXuan Peng · Junteng Tu · Henghui Xiao
Rutgers University · Nanjing Tech University · Nanjing University of Posts and Telecommunications · New York University
Research Digest··2 min read
Bu and colleagues isolate the cost of information loss at interfaces in four-stage LLM pipelines, which they call the decomposition tax.
Why this paper
From Nanjing University of Posts and Telecommunications and 3 others · Part of Context Engineering for Agents, now 41 papers
In one line
Multi-stage LLM pipelines can lose up to 40.5 accuracy points when intermediate stages cannot access the original problem, with re-grounding recovering much of the loss.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§