Artificial Intelligence · arXiv:2606.16613 v2 · submitted 15 Jun 2026
CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies
The paper says
“Under a revenue target, circular trades occur for some pairs of models: one agent proposes that the two rival firms sell goods back and forth at inflated prices so that both book revenue, and the rival accepts.”
from the abstract · picked out as the main claim by a classifier
Reviews
Reviews by people
No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
LLM agents are increasingly capable of handling complex, long-horizon tasks and may eventually take on business management, a role traditionally performed by humans. Deploying them safely requires monitoring their capabilities and behavior in environments that reflect real economies. We introduce CoffeeBench, a benchmark built on a compact simulation of a coffee supply chain with realistic business mechanics. Six autonomous LLM coding agents run its farmers, roasters, and retailers, and negotiate and trade with one another to maximize their own net income over 90 simulated days. We evaluate eight recent proprietary and open-weight models by letting each run one roaster while a fixed background model runs the other five firms, and scoring that roaster's net income. Claude Sonnet 5.5 earns the highest net income, closely followed by GPT-6 Astra. The leaders sell at higher prices, and the best model also buys its inputs far more cheaply, whereas weaker models trade at thin margins. Since performance targets are a common cause of accounting fraud in human firms, we also test whether agents under KPI pressure inflate their numbers. Under a revenue target, circular trades occur for some pairs of models: one agent proposes that the two rival firms sell goods back and forth at inflated prices so that both book revenue, and the rival accepts. We hope CoffeeBench serves as an early step toward a future in which LLM agents manage firms without misconduct. We release our code and agent trajectories.
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-11-16 · no language-model review
How the main claim is stated: asserted 0.90
What the abstract itself states, as probabilities: claims first 0.08 · states limitation 0.41 · human participants 0.03 · controlled comparison 0.59 · randomized assignment 0.03 · states quantitative result 0.80 · compares with prior results 0.70 · mentions code or data availability 0.98
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v2). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| LLM agents are increasingly capable of handling complex, long-horizon tasks and may eventually take on business management, a role traditionally performed by humans. | background 1.00 | 0.05 | 0.03 | hedged possibility 0.94 |
| Deploying them safely requires monitoring their capabilities and behavior in environments that reflect real economies. | background 0.92 | 0.05 | 0.03 | asserted 0.95 |
| We introduce CoffeeBench, a benchmark built on a compact simulation of a coffee supply chain with realistic business mechanics. | contribution 0.93 | 0.93 | 0.04 | asserted 1.00 |
| Six autonomous LLM coding agents run its farmers, roasters, and retailers, and negotiate and trade with one another to maximize their own net income over 90 simulated days. | method 0.95 | 0.88 | 0.04 | asserted 1.00 |
| We evaluate eight recent proprietary and open-weight models by letting each run one roaster while a fixed background model runs the other five firms, and scoring that roaster's net income. | method 0.98 | 0.96 | 0.04 | asserted 0.98 |
| Claude Sonnet 5.5 earns the highest net income, closely followed by GPT-6 Astra. | result 1.00 | 0.96 | 0.34 | asserted 1.00 |
| The leaders sell at higher prices, and the best model also buys its inputs far more cheaply, whereas weaker models trade at thin margins. | result 0.82 | 0.93 | 0.15 | asserted 1.00 |
| Since performance targets are a common cause of accounting fraud in human firms, we also test whether agents under KPI pressure inflate their numbers. | contribution 0.65 | 0.48 | 0.04 | asserted 0.75 |
| Under a revenue target, circular trades occur for some pairs of models: one agent proposes that the two rival firms sell goods back and forth at inflated prices so that both book revenue, and the rival accepts. | result 1.00 | 0.96 | 0.08 | asserted 0.90 |
| We hope CoffeeBench serves as an early step toward a future in which LLM agents manage firms without misconduct. | future work 1.00 | 0.07 | 0.11 | hedged possibility 0.46 |
| We release our code and agent trajectories. | contribution 0.94 | 0.86 | 0.02 | asserted 0.98 |
What kind of paper the classifier reports this is: dataset or benchmark 1.00 · evidence from simulation or computation 0.89
What was examined, and what was not
Sources read
- arXiv listing abstract v2
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-16: 1 request, 8,147 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Cited by 1 Pith paper
Reviewed papers on Pith that cite this one. OpenAlex reports about 12 citations worldwide. The full citation record
How the literature uses it
4 arXiv papers cite this in 4 sentences
Papers citing it · 2 sentences without a confident label
Cites it as background 2 papers
-
A growing line of work has introduced benchmarks that simulate realistic workflows in web environments, software engineering tasks, and office applications, including SWE-bench [ref] , WebArena [ref] , OSWorld [ref] , and more recent workflow-oriented benchmarks such as OdysseyBench [ref] and CoffeeBench [ref] .
PolyWorkBench: Benchmarking LLM Agents for Cross-Lingual Long-Horizon Workflows · Introduction · 0.96
-
Recent long-horizon benchmarks assess sustained reasoning and action across workplace workflows, open-ended exploration, virtual-world planning, web navigation, computer use, inventory control, order fulfillment, and interactive economies [ref] .
MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations · Related Work · 0.86
Challenges and discussion
Know something about this paper? Leave a comment.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment