# Structure for Reading, Prose for Writing: Asymmetric Structural Conditioning in Multi-Agent Document Authoring

arXiv: https://arxiv.org/abs/2608.20786 · HTML: https://mlsanity.com/paper/2608.20786/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Web, search, and research agents, one of 16 directions in LLM agents and tool use.
- Builds on 11 of its 15 references that are in the map.
- First posted on arXiv 21 Aug 2026.

## Placed in

- LLM agents and tool use / [Web, search, and research agents](https://mlsanity.com/agents/web-search-and-research-agents/)

## Builds on (most cited first, 11 of 11)

- [Accelerating scientific discovery with Co-Scientist](https://mlsanity.com/paper/2502.18864/) (2025)
- [Self-Preference Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2410.21819/) (2024)
- [ACON: Optimizing Context Compression for Long-horizon LLM Agents](https://mlsanity.com/paper/2510.00615/) (2025)
- [AgentAsk: Multi-Agent Systems Need to Ask](https://mlsanity.com/paper/2510.07593/) (2025)
- [S-DAG: A Subject-Based Directed Acyclic Graph for Multi-Agent Heterogeneous Reasoning](https://mlsanity.com/paper/2511.06727/) (2025)
- [Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines](https://mlsanity.com/paper/2604.23178/) (2026)
- [The Hallucination Snowball: Modeling Error Propagation as State Transitions in Multi-Agent LLM Pipelines](https://mlsanity.com/paper/2608.14588/) (2026)
- [HELIOS: Hierarchical Graph Abstraction for Structure-Aware LLM Decompilation](https://mlsanity.com/paper/2601.14598/) (2026)
- [You Don't Need Prompt Engineering Anymore: The Prompting Inversion](https://mlsanity.com/paper/2510.22251/) (2025)
- [Questionnaire meets LLM: A Benchmark and Empirical Study of Structural Skills for Understanding Questions and Responses](https://mlsanity.com/paper/2510.26238/) (2025)
- [Compact Constraint Encoding for LLM Code Generation: An Empirical Study of Token Economics and Constraint Compliance](https://mlsanity.com/paper/2604.07192/) (2026)

## Abstract (arXiv)

Multi-agent pipelines that author formal documents must both read a requester's forms and write against them. We report a deployed tender-response system, running an open-weights model under sovereignty constraints, and evaluate it against human-written bids the same organisation actually submitted. On a blind comparison where the system had no worked example available, an LLM judge rated its answers at least as good as the human-submitted answer on 40 of 55 ground-truth sections, better on 4, missing on none, and flagged one unsupported claim in total. Classifying every gap the judge identified shows that 68% were content absent from the system's own sources -- knowledge the human author held and the pipeline was never given -- so only 6 of the 15 adverse verdicts involve a deficiency the system could have avoided. A divergence from ground truth is more often an information-availability result than a writing-quality one, and evaluations that do not separate the two understate such systems. Against this backdrop we report a conditioning asymmetry. It is well established that rendering documents as structural markup rather than flat prose improves extraction, and we reproduce that on three reading tasks. The benefit does not transfer to conditioning: converting a bid's instruction material from prose to nested XML dropped answer quality from 74% to 48% under a paired comparison. We further find that naming a forbidden construction concentrates rather than removes it -- 96% of surviving defects fall in the two forms the prompt explicitly names -- and that coupling a stochastic annotation to a deterministic windowing function moves the extracted requirement count from 68 to 51 on a byte-identical file. Structure belongs where the model reads; prose and self-applied tests belong where it writes.