Pith. sign in

REVIEW 3 major objections 4 minor 40 references

ACWorld claims that agentic commerce can be evaluated reliably only when every agent action is validated before it changes shared market state, with full traces kept so that process evidence — not just final outcomes — decides the score.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 07:14 UTC pith:ZWOI6CDS

load-bearing objection Solid benchmark engineering with real validation, but the central claim of evaluating two independent policies together is not supported by the reported experiments. the 3 major comments →

arxiv 2608.02441 v1 pith:ZWOI6CDS submitted 2026-08-03 cs.AI

Agentic Commerce World: An Auditable and Verifiable Environment for Vibe Commerce

Shicheng Fan , Mingdai Yang , Duohao Wang , Canyu Chen , Yongfeng Zhang , Hua Wei , Manling Li , Julian McAuley
show 4 more authors
Kun Zhang Philip S. Yu Kejing Yu Zhiwei Liu
This is my paper
classification cs.AI
keywords agentic commercevibe commerceLLM agent evaluationmulti-agent environmentprocess-level verificationauditable tracescommerce benchmarkBuyer and Merchant policies
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

ACWorld is built to answer a question existing benchmarks avoid: how should one evaluate Buyer and Merchant agents that share one market, each holding private goals and authority? The paper's answer is to separate deciding from committing — a Vibe Commerce Protocol binds every proposed action to its actor, a platform validates it, and only the world state may change, while the full decision-validation-effect record is kept for every step. The paper then argues, with evidence, that this record is not a luxury: across ten models, 99 of 861 imperfect runs reach the same final state as a fully correct run, so any evaluation that looks only at outcomes misses real errors. The benchmark reports model means of 66–86% on 200 capability tasks and 56–91% on 60 tasks searching a 785,022-listing catalog. If the argument holds, agent evaluation shifts from 'did the job finish' to 'can every step be verified.'

Core claim

The paper's central claim is that an agentic commerce environment can evaluate two independent, privately motivated policies in one persistent market if it separates agent decisions from the authority to change shared state. VCP attaches each proposed action to a named Buyer or Merchant, the Commerce Intelligence Platform accepts or rejects it against market rules, and only the World applies approved events through a deterministic transition; every step leaves a linked record of actor, decision, validation, evidence, and state change. The paper's empirical companion claim is that this process-level evidence is necessary: among 861 capability-coverage runs without full credit, 99 share a fina

What carries the argument

The load-bearing object is the Vibe Commerce Protocol (VCP), a contract layer that binds each action proposal to an authenticated actor and forbids messages from changing world state directly. Its companion is the linked record Lt = (actor, decision, validation, execution evidence, state delta) stored for every evaluated step, plus the deterministic transition W_{j+1} = F(W_j, u_j) through which the World applies only approved events. These records make a run reconstructable from the initial state and rescorable without another model call, which is what turns the paper's scores into verifiable claims rather than model self-reports. The benchmark scoring combines r declared Boolean or fractio

Load-bearing premise

The load-bearing premise is that the benchmark's author-defined scoring predicates correctly capture what counts as a good grounding, negotiation, or purchase decision; if a predicate mis-specifies commercial intent, the reported scores and the 'final state alone misses errors' conclusion inherit that error — a risk the paper itself acknowledges by noting the catalogs have had no independent audit and that reconstruction verifies consistency, not semantic correctness.

What would settle it

Build a second scorer for the same traces using independently authored predicates plus human adjudication of the 99 colliding runs; if the independent rubric finds the process signals add little, or disagrees with the paper's predicate verdicts on a large share of collisions, then the necessity of process-level evidence would be an artifact of the paper's own rubric rather than a property of the environment. A cheaper check: rerun the collision analysis with equal predicate weights and with predicates reordered, and confirm the 99/861 count and the model ordering persist.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Evaluators that score only final outcomes will systematically undercount agent errors; the paper's collision numbers (99/861 for final state, 54/861 for final action, 45/861 for action sequence) quantify that blind spot on this benchmark.
  • Verified intermediate signals — reads, decisions, citations, actions — are candidate process rewards for training commerce agents, since traces rescore deterministically without another model call.
  • Separating decision from authority gives a reproducible protocol boundary: any extension (new product domain, tool, or world event) must pass the same validation and leave enough evidence to reconstruct its effects.
  • Because public tasks invite direct optimization, the paper commits to pairing future public development tasks with unpublished held-out test tasks to keep scores comparable.
  • Capability breadth and catalog-scale execution are distinct dimensions: model ordering differs between the two tracks, and large-catalog bottlenecks concentrate in Search and Decision rather than Validation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The validate-then-commit separation is a general recipe for multi-agent evaluation beyond commerce: any shared-state domain with private actor objectives could adopt the same trace format, and the collision analysis is a template for measuring how much information a restricted evaluator view discards.
  • The 99/861 collision count is likely a lower bound on the blind spot of real deployed evaluators, because it is computed against this benchmark's own predicate verdicts; an independent label set for the same traces would show whether the claim survives outside the author-defined rubric.
  • Because incomplete trajectories retain positive signal on 1.88 of 3.49 predicates on average, a cheap testable extension is to use partial traces directly as training rewards before the full task is solved — the paper explicitly leaves the learning value of these signals unevaluated.
  • The headline scores should not be read as a general leaderboard: the paper frames them as characterizing these tracks, and the design invites re-weightings, new families, and adversarial tasks that would shift the ordering.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper introduces Agentic Commerce World (ACWorld), an environment and benchmark for evaluating LLM-based Buyer and Merchant agents in a shared, persistent commerce market. The core architectural idea is the Vibe Commerce Protocol (VCP), which separates an agent's proposal from the authority to change shared state: the Commerce Intelligence Platform validates each typed action, and only then does the World commit the authorized effect. Every decision is stored as a linked record (actor, decision, validation, execution evidence, state delta), and traces can be deterministically reconstructed and rescored from the initial state and ordered commits. The benchmark has a 200-task capability-coverage track and a 60-task large-catalog track over 785,022 listings. Ten models are evaluated, with mean scores 65.9–85.6% and 56.1–91.4% on the two tracks. The paper also reports a failure-attribution analysis, a restricted-view collision analysis (99/861 runs without full credit share a final state with a full-credit run), a multiagent market study, and a set of validity gates: 200 reference episodes pass, 95 targeted mutations change intended checks, all 2,000 traces reconstruct, and rescoring is deterministic. The authors explicitly limit the scope: each benchmark task evaluates one policy against deterministic counterparties, and the fixed 5×5 study does not support strategy comparisons.

Significance. If the claims hold, the paper makes a useful contribution: it provides a reproducible, auditable environment for agentic commerce, with process-level evidence instead of final-outcome-only scoring, and it backs the environment with unusually thorough internal validation. The strongest parts are the machine-checkable verification infrastructure: all reference episodes pass, targeted mutations affect intended predicates, state reconstruction and deterministic rescoring are verified on all 2,000 traces, and the equal-weight sensitivity analysis (Table 20) checks the hand-set predicate weights. These are concrete, reproducible guarantees that distinguish the work from benchmarks that rely on LLM judges or loose human evaluation. The restricted-view collision finding (Section 5.6) is a well-posed information-insufficiency result, not an overclaimed error-rate statement. The main weakness is that the paper's headline contribution — 'independently controlled Buyer and Merchant policies together' — is not actually exercised by the benchmark protocol, and the one multiagent study uses a single model family and seed. This is a gap between the central claim and the supplied evidence, and it n

major comments (3)
  1. [§1, §4, §5.4, §8] The central claim that ACWorld 'evaluates independently controlled Buyer and Merchant policies together' is not demonstrated by the reported experiments. The benchmark protocol in §4 states that 'Each task assigns one Buyer or Merchant to the evaluated model while every other agent follows a fixed policy.' Thus the 2,000 and 600 runs measure single-policy performance against deterministic scripts, not two independently controlled policies interacting. The multiagent study in §5.4 uses ten agents in a 5×5 market but with one model family and one seed, and §8 admits that 'Neither the scores nor the fixed 5×5 study measure strategy in an open market.' The environment may well support independent two-sided evaluation, but no experiment actually shows two independently controlled policies jointly producing a shared outcome. This is load-bearing because the paper's contribution (ii) and the ab
  2. [§4.2, §8] The benchmark's validity rests on the author-defined task predicates and reference policies. Equation (3) aggregates declared weights w_i over Boolean/fractional predicates c_i, and the entire failure analysis treats these predicates as ground truth. The paper's internal validation (reference episodes pass, mutations lower intended checks, deterministic rescoring) demonstrates self-consistency but not that the predicates correctly operationalize commercial intent. The paper itself notes in §8 that 'Reconstruction establishes consistency under the declared transition and scorer, not their independent semantic correctness' and that catalogs have not received an independent audit. Given that the benchmark is a central deliverable, the absence of any external check — e.g., independent human annotation of a sample of predicate verdicts, or a comparison with a practitioner-defined scoring rubr
  3. [§5.4, Appendix K] The multiagent environment study is too thinly specified to support the 'independent policies in a shared market' narrative. The 5×5 study uses one model family, one seed, and a single deterministic clearing rule; the sparse-world probe in Appendix K uses scripted agents and excludes model inference, VCP routing, and contention. The paper appropriately disclaims that 'One market, model family, and seed cannot support strategy comparisons,' but this disclaimer also means the study provides only existence evidence that the environment can execute a multiagent market. Concretely, I would like to see either (i) repeated runs over multiple seeds with the current model, and (ii) at least one configuration where the Buyer and Merchant roles are played by different model families. Without this, the unique two-sided claim remains an architectural promise rather than an empirically supported prope
minor comments (4)
  1. [§5.5.2 / Figure 9] The case study gives a clear example of partial credit, but the weight values (0.214, 0.286, etc.) appear without explaining how the four predicates sum to 1.0. A one-line note that weights are normalized per task would help.
  2. [§5.6, Finding 3] Finding 3 is explicitly qualitative and based on 'qualitative inspection.' It is appropriately labeled as an observation, but the text could more clearly state that no inter-annotator agreement or statistical test was used, to avoid readers treating it as a measured model property.
  3. [Table 3 / Table 7] The skill manifests are counted as 11 Buyer and 25 Merchant in Table 3, and Table 7 lists the same names grouped by function. It would improve readability to note that a skill manifest can have multiple primary functions, or that the counts refer to manifests rather than individual skills, since the grouping in Table 7 does not map one-to-one to the numbers.
  4. [Appendix D] Table 11 shows that 17 of 95 targeted mutations are not detectable from final state, but the main text's Finding 1 uses the empirical collision rate of 11.5% from Table 5. It would be helpful to connect these two numbers explicitly: the mutation result is a designed lower bound on blind spots, while the collision result measures observed frequency in the model panel.

Circularity Check

0 steps flagged

No material circularity: scoring is declared-predicate aggregation, collision analysis is an explicit information-theoretic projection, and self-citations are motivational only.

full rationale

The paper's evaluation is not a fitted prediction or a derivation that reduces to its own input. Scores are computed by Eq. (3) from declared task predicates whose weights are hand-set and explicitly sensitivity-checked (Table 20), and the 'full trace' evidence record (Eq. (2)) is defined independently of any model output. Finding 1's restricted-view collisions (Section 5.6, Table 5) are presented precisely as an information-insufficiency calculation over saved projections: 'This establishes information insufficiency, not the error rate of a particular evaluator.' The 99/861 count is therefore a logical consequence of the defined projections, not a fitted result dressed as a prediction. The 'verifiable' language is also explicitly bounded: 'verifiable means reconstructable and rescorable under the declared ACWORLD contract, not proven transition semantics or production security.' Self-citations (Fan et al. 2026a; Zhang et al. 2026b/c) appear only as pointers to future policy-learning uses of process rewards, not as load-bearing support for the benchmark's validity. That the same team designed the tasks, predicates, and reference policies is ordinary benchmark engineering, and Section 8 candidly states the limits: 'Each task evaluates one policy against deterministic counterparties.' No circular step satisfying the evidence threshold was found.

Axiom & Free-Parameter Ledger

1 free parameters · 3 axioms · 3 invented entities

The paper contributes an engineered evaluation environment rather than a mathematical derivation. The main free parameters are the hand-set predicate weights. The axioms are domain assumptions about deterministic worlds and author-defined scoring correctness. No new physical or scientific entities are postulated; the invented entities are software components with released implementations.

free parameters (1)
  • Task predicate weights w_i = 0.05 to 1.0, normalized within each task
    Hand-declared per predicate for scoring in Eq. (3). Scores depend on these weights, though Table 20 shows equal-weight sensitivity preserves model ordering.
axioms (3)
  • domain assumption The World transition W_{j+1} = F(W_j, u_j) is deterministic and fully captures all commercial effects of accepted actions.
    Invoked in §3.1; state reconstruction and rescoring depend on this determinism.
  • ad hoc to paper Programmatic predicates and declared weights w_i in Eq. (3) correctly operationalize each task's commercial intent without an LLM judge.
    §4.2; the validity of the benchmark's central scoring and failure analysis rests on author-defined predicates. Sensitivity is checked but semantic correctness is not externally audited.
  • domain assumption Counterparty policies are fixed and deterministic for controlled comparison.
    §5.1 and §8 state that counterparties and environment behavior are deterministic; the results therefore characterize behavior against scripts, not open-market strategy.
invented entities (3)
  • Vibe Commerce Protocol (VCP) independent evidence
    purpose: Binds each proposed action to a Buyer or Merchant actor and separates proposal from authorization and World effects, enabling auditability.
    Modern commerce platform; implemented and released in the cited repository, with concrete message schemas in Fig. 2.
  • Commerce Intelligence Platform independent evidence
    purpose: Central authority that validates agent actions against rules and world state before any state change is committed.
    Software component with declared validation rules and policy engine; implementation is part of the released repository.
  • ACWorld persistent World independent evidence
    purpose: Shared source of truth for inventory, payment, and trust records, updated only through deterministic atomic commits.
    The environment itself; its reconstruction and digest checks are implemented and reproducible through the repository.

pith-pipeline@v1.3.0-daily-deepseek · 25955 in / 11774 out tokens · 122832 ms · 2026-08-04T07:14:36.500300+00:00 · methodology

0 comments
read the original abstract

In vibe coding, people describe software in natural language and delegate implementation to AI agents. By analogy, vibe commerce allows people to express buying or selling goals in natural language and delegate the corresponding tasks to agents. Commerce, however, requires independently controlled Buyer and Merchant agents to interact in a shared market while preserving their private objectives and distinct authority. We introduce Agentic Commerce World (ACWorld), an environment for evaluating such agents across ongoing transactions. Through its Vibe Commerce Protocol (VCP), ACWorld validates agent actions before updating shared transaction state and records the resulting interactions, making agent behavior auditable and evaluation reproducible. The ACWorld Benchmark contains a 200-task capability-coverage track and a 60-task large-catalog track that searches 785,022 transactable listings. Across ten models, mean scores range from 65.9% to 85.6% and from 56.1% to 91.4%, respectively. Our analysis shows that process-level evidence is necessary: final state alone can miss evaluated errors, incomplete trajectories still retain useful process signals, and large-catalog tasks expose bottlenecks across stages.

Figures

Figures reproduced from arXiv: 2608.02441 by Canyu Chen, Duohao Wang, Hua Wei, Julian McAuley, Kejing Yu, Kun Zhang, Manling Li, Mingdai Yang, Philip S. Yu, Shicheng Fan, Yongfeng Zhang, Zhiwei Liu.

Figure 1
Figure 1. Figure 1: A vibe-commerce negotiation. A Buyer with a private $100 ceiling and a Merchant with a [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Two VCP messages. Buyer b1 offers $80, and Merchant m1 counters at $100. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: ACWORLD execution path for a $95 grinder purchase. Skills help each role produce VCP messages. The Commerce Intelligence Platform validates them and commits authorized effects to the World. The diagram follows this path from Buyer and Merchant skills through search, offer, and purchase commit to shared inventory, payment, and trust records. VCP enforces four invariants. Each action belongs to an authentica… view at source ↗
Figure 4
Figure 4. Figure 4: Grounding task with feedback linked to evidence. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Benchmark composition. The inner rings show the role and family structure of capability [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Credit distributions for each model on both benchmark tracks. Full credit means that every [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Capability-coverage mean score (%) by task family for the ten evaluated models. Exact [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Failure composition by task family. The right column gives each family’s share of runs [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: A real partial-credit trace. The Buyer makes a valid purchase but chooses cheaper Item A [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Execution trace for the running negotiation example. [PITH_FULL_IMAGE:figures/full_fig_p031_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Detection of 95 injected errors under four evidence views. Filled cells are detected, and [PITH_FULL_IMAGE:figures/full_fig_p032_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Sparse World-only component probe over three fixed seeds per size. The workload mate [PITH_FULL_IMAGE:figures/full_fig_p033_12.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 1 canonical work pages · 1 internal anchor

  1. [1]

    Advances in Neural Information Processing Systems , volume =

    Yao, Shunyu and Chen, Howard and Yang, John and Narasimhan, Karthik , title =. Advances in Neural Information Processing Systems , volume =. 2022 , url =

    2022

  2. [2]

    International Conference on Learning Representations , volume=

    Webarena: A realistic web environment for building autonomous agents , author=. International Conference on Learning Representations , volume=

  3. [3]

    International Conference on Learning Representations , editor =

    Yao, Shunyu and Shinn, Noah and Razavi, Pedram and Narasimhan, Karthik , title =. International Conference on Learning Representations , editor =. 2025 , url =

    2025

  4. [4]

    Advances in Neural Information Processing Systems , volume =

    Deng, Xiang and Gu, Yu and Zheng, Boyuan and Chen, Shijie and Stevens, Samuel and Wang, Boshi and Sun, Huan and Su, Yu , title =. Advances in Neural Information Processing Systems , volume =. 2023 , publisher =

    2023

  5. [5]

    International Conference on Learning Representations , year =

    Liu, Xiao and Yu, Hao and Zhang, Hanchen and Xu, Yifan and Lei, Xuanyu and Lai, Hanyu and Gu, Yu and Ding, Hangliang and Men, Kaiwen and Yang, Kejuan and Zhang, Shudan and Deng, Xiang and Zeng, Aohan and Du, Zhengxiao and Zhang, Chenhui and Shen, Sheng and Zhang, Tianjun and Su, Yu and Sun, Huan and Huang, Minlie and Dong, Yuxiao and Tang, Jie , title =. ...

  6. [6]

    International Conference on Learning Representations , year =

    Mialon, Gr. International Conference on Learning Representations , year =

  7. [7]

    2026 , publisher=

    WebMall-A multi-shop benchmark for evaluating web agents , author=. 2026 , publisher=

    2026

  8. [8]

    Findings of the Association for Computational Linguistics: NAACL 2025 , year =

    Lu, Jiarui and Holleis, Thomas and Zhang, Yizhe and Aumayer, Bernhard and Nan, Feng and Bai, Haoping and Ma, Shuang and Ma, Shen and Li, Mengyu and Yin, Guoli and Wang, Zirui and Pang, Ruoming , title =. Findings of the Association for Computational Linguistics: NAACL 2025 , year =. doi:10.18653/v1/2025.findings-naacl.65 , url =

    doi:10.18653/v1/2025.findings-naacl.65 2025

  9. [9]

    2026 , eprint =

    Liu, Xianyang and Gu, Shangding and Song, Dawn , title =. 2026 , eprint =. doi:10.48550/arXiv.2602.06008 , url =

    doi:10.48550/arxiv.2602.06008 2026

  10. [10]

    arXiv preprint arXiv:2510.25779 , year=

    Magentic Marketplace: An Open-Source Environment for Studying Agentic Markets , author=. arXiv preprint arXiv:2510.25779 , year=

    arXiv

  11. [11]

    Proceedings of the Natural Legal Language Processing Workshop 2025 , pages =

    Zhu, Shenzhe and Sun, Jiao and Nian, Yi and South, Tobin and Pentland, Alex and Pei, Jiaxin , title =. Proceedings of the Natural Legal Language Processing Workshop 2025 , pages =. 2025 , doi =

    2025

  12. [12]

    Advances in Neural Information Processing Systems , volume =

    Debenedetti, Edoardo and Zhang, Jie and Balunovi. Advances in Neural Information Processing Systems , volume =. 2024 , numpages =. doi:10.52202/079017-2636 , url =

    doi:10.52202/079017-2636 2024

  13. [13]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    Wang, Jiangyuan and Xiao, Kejun and Sun, Qi and Zhao, Huaipeng and Luo, Tao and Zhang, Jian Dong and Zeng, Xiaoyi , title =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

    2026

  14. [14]

    arXiv preprint arXiv:2606.16613 , year=

    CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies , author=. arXiv preprint arXiv:2606.16613 , year=

    arXiv

  15. [15]

    Terry, J. K. and Black, Benjamin and Grammel, Nathaniel and Jayakumar, Mario and Hari, Ananth and Sullivan, Ryan and Santos, Luis S. and Dieffendahl, Clemens and Horsch, Caroline and Perez-Vicente, Rodrigo and Williams, Niall and Lokesh, Yashas and Ravi, Praveen , title =. Advances in Neural Information Processing Systems , volume =. 2021 , url =

    2021

  16. [16]

    Leibo, Joel Z. and Due. Scalable Evaluation of Multi-Agent Reinforcement Learning with. Proceedings of the 38th International Conference on Machine Learning , series =. 2021 , url =

    2021

  17. [17]

    arXiv preprint arXiv:2312.03664 , year=

    Generative agent-based modeling with actions grounded in physical, social, or digital space using Concordia , author=. arXiv preprint arXiv:2312.03664 , year=

    Pith/arXiv arXiv

  18. [18]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics , pages =

    Zhu, Kunlun and Du, Hongyi and Hong, Zhaochen and Yang, Xiaocheng and Guo, Shuyi and Wang, Zhe and Wang, Zhenhailong and Qian, Cheng and Tang, Xiangru and Ji, Heng and You, Jiaxuan , title =. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics , pages =. 2025 , doi =

    2025

  19. [19]

    arXiv preprint arXiv:2605.16116 , year=

    ShopGym: An Integrated Framework for Realistic Simulation and Scalable Benchmarking of E-Commerce Web Agents , author=. arXiv preprint arXiv:2605.16116 , year=

    Pith/arXiv arXiv

  20. [20]

    What Is Your

    Allouah, Amine and Besbes, Omar and Figueroa, Josu. What Is Your. Proceedings of the ACM Web Conference 2026 , pages =. 2026 , doi =

    2026

  21. [21]

    arXiv preprint arXiv:2606.17698 , year=

    EComAgentBench: Benchmarking Shopping Agents on Long-Horizon Tasks with Distributed Hidden Intent , author=. arXiv preprint arXiv:2606.17698 , year=

    Pith/arXiv arXiv

  22. [22]

    arXiv preprint arXiv:2606.12608 , year=

    Shopping Reasoning Bench: An Expert-Authored Benchmark for Multi-Turn Conversational Shopping Assistants , author=. arXiv preprint arXiv:2606.12608 , year=

    Pith/arXiv arXiv

  23. [23]

    arXiv preprint arXiv:2603.16453 , year=

    RetailBench: Evaluating Long-Horizon Autonomous Decision-Making and Strategy Stability of LLM Agents in Realistic Retail Environments , author=. arXiv preprint arXiv:2603.16453 , year=

    Pith/arXiv arXiv

  24. [24]

    Proceedings of the 43rd International Conference on Machine Learning , series =

    Barres, Victor and Dong, Honghua and Ray, Soham and Si, Xujie and Narasimhan, Karthik , title =. Proceedings of the 43rd International Conference on Machine Learning , series =. 2026 , url =

    2026

  25. [25]

    Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , year =

    Trivedi, Harsh and Khot, Tushar and Hartmann, Mareike and Manku, Ruskin and Dong, Vinty and Li, Edward and Gupta, Shashank and Sabharwal, Ashish and Balasubramanian, Niranjan , title =. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , year =. doi:10.18653/v1/2024.acl-long.850 , url =

    doi:10.18653/v1/2024.acl-long.850 2024

  26. [26]

    Advances in Neural Information Processing Systems 37 , year =

    Xie, Tianbao and Zhang, Danyang and Chen, Jixuan and Li, Xiaochuan and Zhao, Siheng and Cao, Ruisheng and Hua, Toh Jing and Cheng, Zhoujun and Shin, Dongchan and Lei, Fangyu and Liu, Yitao and Xu, Yiheng and Zhou, Shuyan and Savarese, Silvio and Xiong, Caiming and Zhong, Victor and Yu, Tao , title =. Advances in Neural Information Processing Systems 37 , ...

    doi:10.52202/079017-1650

  27. [27]

    arXiv preprint arXiv:2605.10059 , year=

    Strategic Exploitation in LLM Agent Markets: A Simulation Framework for E-Commerce Trust , author=. arXiv preprint arXiv:2605.10059 , year=

    Pith/arXiv arXiv

  28. [28]

    arXiv preprint arXiv:2605.13909 , year=

    TERMS-Bench: Diagnosing LLM Negotiation Agents Beyond Deal Rate , author=. arXiv preprint arXiv:2605.13909 , year=

    Pith/arXiv arXiv

  29. [29]

    arXiv preprint arXiv:2606.08790 , year=

    RAILS: Verification-Native Clearing For Agentic Commerce , author=. arXiv preprint arXiv:2606.08790 , year=

    Pith/arXiv arXiv

  30. [30]

    A Decision-Centered Reference Architecture for Trustworthy Agentic Commerce

    Sfiris, Dimitrios S. , title =. 2026 , eprint =. doi:10.48550/arXiv.2607.18347 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2607.18347 2026

  31. [31]

    Agentic Commerce Protocol , year =

  32. [32]

    Universal Commerce Protocol , year =

  33. [33]

    Agent Payments Protocol , year =

  34. [34]

    Shopify Agentic Storefronts , year =

  35. [35]

    2026 , howpublished =

    Models: One. 2026 , howpublished =

    2026

  36. [36]

    2026 , howpublished =

    2026

  37. [37]

    2026 , howpublished =

    Introducing. 2026 , howpublished =

    2026

  38. [38]

    and Liu, Xue , title =

    Zhang, Hanrong and Chen, Yankai and Fan, Shicheng and Min, Dehai and Chen, Shaowen and Ma, Huanhuan and Wu, Zhaofen and Yang, Jie and He, Bowei and Kang, Jikun and Zheng, Kening and Chen, Xi and Miao, Chunyu and Lin, Fulin and Huang, Wei-Chieh and Zhou, Jiayu and Wu, Haolun and Fang, Liancheng and Kang, Hong and He, Langzhou and Zou, Henry Peng and Li, Ch...

    2026

  39. [39]

    and Cheng, Lu , title =

    Fan, Shicheng and Hao, Haochang and Min, Dehai and Liu, Weihao and Yu, Philip S. and Cheng, Lu , title =. arXiv preprint arXiv:2605.29648 , year =

    Pith/arXiv arXiv

  40. [40]

    arXiv preprint arXiv:2604.01687 , year =

    Zhang, Hanrong and Fan, Shicheng and Zou, Henry Peng and Chen, Yankai and Wang, Zhenting and Zhou, Jiayu and Li, Chengze and Huang, Wei-Chieh and Yao, Yifei and Zheng, Kening and others , title =. arXiv preprint arXiv:2604.01687 , year =

    Pith/arXiv arXiv