# Floor, Ceiling, and the Fusion Gap: How Much of Crowd Reading Attention Can Machines Predict?

arXiv: https://arxiv.org/abs/2608.01704 · HTML: https://mlsanity.com/paper/2608.01704/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 11 of its 12 references that are in the map.
- First posted on arXiv 3 Aug 2026.

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

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference](https://mlsanity.com/paper/2412.13663/) (2024)
- [LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression](https://mlsanity.com/paper/2403.12968/) (2024)
- [Great Models Think Alike and this Undermines AI Oversight](https://mlsanity.com/paper/2502.04313/) (2025)
- [Consensus is Not Verification: Why Crowd Wisdom Strategies Fail for LLM Truthfulness](https://mlsanity.com/paper/2603.06612/) (2026)
- [Wisdom from Diversity: Bias Mitigation Through Hybrid Human-LLM Crowds](https://mlsanity.com/paper/2505.12349/) (2025)
- [Personal Salience: Highlighting Is Social, but Individuality Lives in Selection](https://mlsanity.com/paper/2606.09024/) (2026)
- [The Law of Multi-Model Collaboration: Scaling Limits of Model Ensembling for Large Language Models](https://mlsanity.com/paper/2512.23340/) (2025)
- [The Long Tail, Not the Front Page: Cold-Start Prediction of Crowd Highlight Salience](https://mlsanity.com/paper/2606.11654/) (2026)
- [Language Models Agree With Each Other, Not With Readers](https://mlsanity.com/paper/2607.29274/) (2026)

## Built on by (most cited first, 1 of 1)

- [More Reader Data or a Better Teacher? Updating a Model of Crowd Highlights in Late 2026](https://mlsanity.com/paper/2610.17768/) (2026)

## Abstract (arXiv)

A benchmark score means nothing without knowing what a trivial method achieves and what the best possible method could achieve. We construct both bounds for a task with a rare kind of ground truth: predicting which sentences a crowd of readers -- highlighting for their own purposes, unpaid, uninstructed, and blind to each other -- marked in 120 web documents. The floor is naive truncation (lead); the ceiling is a split-half oracle: half the crowd predicting the other half. The gap between them is +0.2028 AP [+0.1698, +0.2342, domain-clustered], and three findings structure it. First, the gap is semantic: position and length features recover 5% of it. Second, frontier language models reach 35-53% of it zero-shot -- far above classical baselines, far below the crowd; a state-of-the-art prompt compressor (LLMLingua-2) lands below the floor, indistinguishable from random selection. Third, an unweighted cross-vendor fusion of five frontier rankings plus a position prior reaches 60%, beating the best single model by +0.0159 [+0.0044, +0.0269; Holm p=0.019] -- a gain that survives ablation of its best member, split-half arm selection, prompt paraphrase, and label, gate, and seed perturbations, and was CONFIRMED by a pre-registered replication on 217 independent documents (+0.0179, Holm p=0.042). Finally, the bracket compresses: distilling the fusion into one open-weight 8B student that reads the whole document retains 90% of the fusion's edge and reaches statistical parity with the strongest single frontier model (+0.0070 [-0.0068, +0.0200]), where a local-context student retains only 63% -- the crowd's signal lives in document-level structure, and the cheapest known improvement is to ask several different models and average.