# The Cost of Knowing: A Resource-Aware Protocol for Benchmarking Hallucination Beyond Static Leaderboards

arXiv: https://arxiv.org/abs/2607.24063 · HTML: https://mlsanity.com/paper/2607.24063/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 43 references that are in the map.
- First posted on arXiv 27 Jul 2026.

## Builds on (most cited first, 22 of 22)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework](https://mlsanity.com/paper/2308.00352/) (2023)
- [DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models](https://mlsanity.com/paper/2309.03883/) (2023)
- [Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems](https://mlsanity.com/paper/2504.01990/) (2025)
- [AgentCF: Collaborative Learning with Autonomous Language Agents for Recommender Systems](https://mlsanity.com/paper/2310.09233/) (2023)
- [Can Large Language Models Serve as Rational Players in Game Theory? A Systematic Analysis](https://mlsanity.com/paper/2312.05488/) (2023)
- [The Factual Inconsistency Problem in Abstractive Text Summarization: A Survey](https://mlsanity.com/paper/2104.14839/) (2021)
- [In-context Pretraining: Language Modeling Beyond Document Boundaries](https://mlsanity.com/paper/2310.10638/) (2023)
- [The Leaderboard Illusion](https://mlsanity.com/paper/2504.20879/) (2025)
- [MechAgents: Large language model multi-agent collaborations can solve mechanics problems, generate new data, and integrate knowledge](https://mlsanity.com/paper/2311.08166/) (2023)
- [Large Language Model Enhanced Multi-Agent Systems for 6G Communications](https://mlsanity.com/paper/2312.07850/) (2023)
- [Deciphering Digital Detectives: Understanding LLM Behaviors and Capabilities in Multi-Agent Mystery Games](https://mlsanity.com/paper/2312.00746/) (2023)
- [AgencyBench: Benchmarking the Frontiers of Autonomous Agents in 1M-Token Real-World Contexts](https://mlsanity.com/paper/2601.11044/) (2026)
- [BatGPT: A Bidirectional Autoregessive Talker from Generative Pre-trained Transformer](https://mlsanity.com/paper/2307.00360/) (2023)
- [Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards](https://mlsanity.com/paper/2501.07493/) (2025)
- [InnovatorBench: Evaluating Agents' Ability to Conduct Innovative LLM Research](https://mlsanity.com/paper/2510.27598/) (2025)
- [LIMI: Less is More for Agency](https://mlsanity.com/paper/2509.17567/) (2025)
- [Neural Knowledge Bank for Pretrained Transformers](https://mlsanity.com/paper/2208.00399/) (2022)
- [DatasetResearch: Benchmarking Agent Systems for Demand-Driven Dataset Discovery](https://mlsanity.com/paper/2508.06960/) (2025)
- [daVinci-Agency: Unlocking Long-Horizon Agency Data-Efficiently](https://mlsanity.com/paper/2602.02619/) (2026)

## Abstract (arXiv)

On standard factuality tasks, frontier models now cluster near the top of the scale. The question is therefore shifting from how factual a system is toward how much compute that factuality costs. Static leaderboards score factuality in isolation and treat compute as free, so they cannot tell a genuinely better system apart from one that simply spends more. Consider a ranking reversal. A brute-force Best-of-4 agent posts the higher raw factuality score (H-Score 0.9169 vs 0.9103) and would top a static leaderboard, but once cost is counted it is the worse system, losing on Q-Score (0.5169 vs 0.5217) at roughly four times the tokens and latency, under a reported cost weight whose sensitivity we sweep. So the system that tops a static leaderboard can be the worse one to deploy. To make this trade-off visible, we introduce MAS-HQ (Multi-Agent System Hallucination Quest), a resource-aware evaluation protocol. It wraps any factuality detector and normalizes for cost, and it pits systems against each other rather than scoring them in isolation. The Q-Score measures factuality minus normalized cost under a competitive match. Across summarization and open-domain QA, single-agent baselines drift into resource-heavy over-optimization, while competition elicits more resource-efficient policies. These gains are small but consistent, and stable across 100 trials. The axis stays discriminative for frontier systems (Gemini-2.5-Pro, and GPT-5) whose raw factuality scores are already bunched near the ceiling. MAS-HQ provides a reproducible way to measure how much a factual answer costs.