Instruction Stacking Collapse: A Benchmark and the Capability-Dependent Value of Prompt Compilation
Source
Atul Anand, Sourav Chattaraj
Author
Atul Anand, Sourav Chattaraj
Date
Key takeaways · AI-distilled
Stacking verifier-checked instructions (1 to 20 at a time, from a pool of 24) degrades follow rate non-linearly: from ~96% down to as low as 20% across Claude Sonnet 4.6, GPT-5-mini and Gemini 2.5 Flash.
The damage is structured, not random: reproducible pairwise conflicts drive the collapse - a single output-JSON constraint is jointly unsatisfiable with nine of the other instructions.
A training-free instruction compiler (one LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition → call that rewrites the stacked prompt, reused across queries) recovers up to +11 points of follow rate - but only for weaker models; stronger models already internalise the structure.
Controls attribute the gain to the rewrite itself, not extra tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →, reordering or measurement headroom - so prompt compilation is worth it exactly where cheap models are deployed at scale.
Terms in this piece · Glossary
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
Why it matters
Stacking many instructions in a single prompt causes non-linear, reproducible failure (e.g., JSON output conflicting with other constraints); a training-free prompt-compiler rewrite recovers up to 11 points of instruction-following for weaker models, offering a cheap mitigation for production prompt pipelines.
Key quotes
“Instruction-following degrades non-linearly: the follow rate falls from ~96% to as low as 20%, driven by a structured and reproducible set of pairwise conflicts.”
“A single "output JSON" constraint, for example, is jointly unsatisfiable with nine others.”
“It recovers up to +11 points of follow rate for weaker models, which are also the models most often deployed at scale, while leaving stronger models, which already internalise the same structure, essentially unchanged.”
articleCompound Prompt Constraints in LLM Code Generation: A Factorial Study of Format, Persona, and UrgencyShrenik Jadhav, Nickalsa LaPlaca, Caleb Stone, Ashok Raja, Omar Ochoa, Vidhyashree Nagaraju
articleSteering Instruction Hierarchies at Inference TimeSiqi Zeng, Sewoong Lee, Han Zhao, Julia Hockenmaier
articleHow well LLM-based test generation techniques perform with newer LLM versions?Michael Konstantinou, Renzo Degiovanni, Mike Papadakis