pith. sign in

arxiv: 2604.20732 · v1 · submitted 2026-04-22 · 💻 cs.MA · cs.AI· cs.CL

Anchor-and-Resume Concession Under Dynamic Pricing for LLM-Augmented Freight Negotiation

Hoang Nguyen , Lu Wang , Marta Gaia Bras This is my paper

Pith reviewed 2026-05-09 22:44 UTC · model grok-4.3

classification 💻 cs.MA cs.AIcs.CL
keywords pricingbetaanchor-and-resumeconcessionratessavingsunderagreement
0
0 comments X

The pith

Anchor-and-resume with spread-derived beta allows adaptive monotonic concessions in freight negotiations, achieving LLM-like performance with lower cost and higher transparency.

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

In freight brokering, prices change often during talks with carriers. Traditional methods use a fixed beta for how much to concede over time, but this doesn't work when targets update. The new method calculates beta from the current margin spread to decide the concession style. To prevent offers from going back when prices change, it uses an anchor point and resumes from there. This keeps everything in simple formulas. The large language model is only used to turn the numbers into natural language messages. Tests with over one hundred thousand simulated negotiations show that in tight price spreads it concedes fast to close deals, and in wider spreads it saves money well. It performs similarly to a big 20 billion parameter LLM but can handle many negotiations at once without high computing costs.

Core claim

We propose a two-index anchor-and-resume framework that addresses both limitations. A spread-derived β maps each load's margin structure to the correct concession posture, while the anchor-and-resume mechanism guarantees monotonically non-decreasing offers under arbitrary pricing shifts. All pricing decisions remain in a deterministic formula; the LLM, when used, serves only as a natural-language translation layer. Empirical evaluation across 115,125 negotiations shows that the adaptive β tailors behavior by regime: in narrow spreads, it concedes quickly to prioritize deal closure and load coverage; in medium and wide spreads, it matches or exceeds the best fixed-β baselines in broker savings.

Load-bearing premise

The assumption that a spread-derived β can be computed in a way that correctly adapts to different regimes without post-hoc adjustments, and that the anchor-and-resume mechanism fully prevents retraction of offers under any pricing shift while preserving the intended concession behavior.

Figures

Figures reproduced from arXiv: 2604.20732 by Hoang Nguyen, Lu Wang, Marta Gaia Bras.

Figure 1
Figure 1. Figure 1: Carrier archetype concession curves. Dashed lines mark 𝑟min and 𝑟max. Annotations in￾dicate walk-away zones for Hardliner (round 8) and Anchoring (round 9). at 95% of the range, makes one large initial drop, then con￾cedes only 2% per round; it walks away after round 9 if the broker remains below 50% of the range [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Rule-based evaluation: 105,000 negotiations (21,000 per strategy). (a) By carrier archetype, with [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Sensitivity of the two-index strategy to calibration constant [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Two-Index (𝑐 = 3, green) vs. unconstrained LLM broker (purple) vs. Two-Index against LLM-powered carriers (red). (a) By carrier archetype, with Overall column. (b) By spread regime. LLM carriers, the two-index strategy achieves 75.8% agree￾ment (±1.0), 0.707 savings (±0.031), and 6.4 rounds (±0.1), with zero retractions across all 6,750 negotiations. Agree￾ment is significantly higher than against algorith… view at source ↗
Figure 5
Figure 5. Figure 5: Offer curves for all five strategies against each carrier archetype (rows: Tit-for-Tat, Hardliner, [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
read the original abstract

Freight brokerages negotiate thousands of carrier rates daily under dynamic pricing conditions where models frequently revise targets mid-conversation. Classical time-dependent concession frameworks use a fixed shape parameter $\beta$ that cannot adapt to these updates. Deriving $\beta$ from the live spread enables adaptation but introduces a new problem: a pricing shift can cause the formula to retract a previous offer, violating monotonicity. LLM-powered brokers offer flexibility but require expensive reasoning models, produce non-deterministic pricing, and remain vulnerable to prompt injection. We propose a two-index anchor-and-resume framework that addresses both limitations. A spread-derived $\beta$ maps each load's margin structure to the correct concession posture, while the anchor-and-resume mechanism guarantees monotonically non-decreasing offers under arbitrary pricing shifts. All pricing decisions remain in a deterministic formula; the LLM, when used, serves only as a natural-language translation layer. Empirical evaluation across 115,125 negotiations shows that the adaptive $\beta$ tailors behavior by regime: in narrow spreads, it concedes quickly to prioritize deal closure and load coverage; in medium and wide spreads, it matches or exceeds the best fixed-$\beta$ baselines in broker savings. Against an unconstrained 20-billion-parameter LLM broker, it achieves similar agreement rates and savings. Against LLM-powered carriers as more realistic stochastic counterparties, it maintains comparable savings and higher agreement rates than against rule-based opponents. By decoupling the LLM from pricing logic, the framework scales horizontally to thousands of concurrent negotiations with negligible inference cost and transparent decision-making.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a two-index anchor-and-resume framework for concession strategies in freight negotiations under dynamic pricing. A spread-derived β adapts the concession posture to the load's margin structure (quick concession in narrow spreads for closure, slower in medium/wide for savings), while the anchor-and-resume mechanism ensures offers remain monotonically non-decreasing despite arbitrary target price revisions. All core pricing uses deterministic formulas; the LLM is restricted to natural-language translation. Evaluation across 115,125 negotiations shows the adaptive β outperforms or matches fixed-β baselines by regime and achieves comparable agreement rates and savings to an unconstrained 20B-parameter LLM broker, with higher agreement rates against LLM-powered stochastic carriers than rule-based ones.

Significance. If the central claims hold, the work provides a practical, scalable, and transparent method for adaptive negotiation that combines deterministic pricing logic with LLM assistance without exposing pricing to non-determinism or prompt injection. The large-scale empirical evaluation (115k+ negotiations) and explicit decoupling of LLM from pricing decisions are notable strengths, enabling horizontal scaling with negligible inference cost. The regime-specific adaptation via spread-derived β addresses a clear limitation of fixed-β classical models. However, significance depends on resolving the monotonicity preservation under regime-crossing shifts.

major comments (2)
  1. [Abstract and framework description] Abstract and framework description: the claim that the anchor-and-resume mechanism 'guarantees monotonically non-decreasing offers under arbitrary pricing shifts' while preserving the intended concession posture from spread-derived β lacks a derivation showing that the resume override (when β jumps regimes, e.g., wide-spread slow-concession to narrow-spread fast-concession) keeps the overall concession curve inside the intended family of functions for every sequence of shifts. This is load-bearing for the adaptation-without-retraction central claim.
  2. [Empirical evaluation section] Empirical evaluation section: aggregate statistics are reported across 115,125 negotiations, but the subset of trajectories crossing regime boundaries mid-negotiation is not isolated or analyzed. Without this, the results cannot confirm that the mechanism preserves the claimed regime-specific concession behaviors (quick closure in narrow spreads, savings-matching in wider) under live β recomputation.
minor comments (2)
  1. The two indices in the 'two-index anchor-and-resume framework' and the exact functional form of the spread-derived β (including any free parameters in its derivation) should be presented with explicit equations or pseudocode to support reproducibility.
  2. Additional details are needed on experimental design: how LLM counterparties were implemented (prompts, models, stochasticity), exact baselines (which fixed-β values, how chosen), negotiation simulation protocol, and precise definitions of 'broker savings' and 'agreement rates'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments highlight important aspects of our monotonicity claim and empirical validation that we will strengthen in revision. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Abstract and framework description] Abstract and framework description: the claim that the anchor-and-resume mechanism 'guarantees monotonically non-decreasing offers under arbitrary pricing shifts' while preserving the intended concession posture from spread-derived β lacks a derivation showing that the resume override (when β jumps regimes, e.g., wide-spread slow-concession to narrow-spread fast-concession) keeps the overall concession curve inside the intended family of functions for every sequence of shifts. This is load-bearing for the adaptation-without-retraction central claim.

    Authors: We agree that an explicit derivation covering regime transitions is needed to fully support the central claim. The anchor-and-resume mechanism prevents retraction by always taking the maximum of the prior offer and the new β-adjusted target, then resuming the concession schedule from that point. However, we acknowledge that the manuscript presents this at a conceptual level without a formal inductive argument for arbitrary sequences of β jumps. In the revision we will add a lemma and proof in Section 3 demonstrating that, for any sequence of regime shifts, offers remain monotonically non-decreasing and each segment follows the concession family dictated by the current β (quick closure for narrow spreads, savings-oriented for wider). revision: yes

  2. Referee: [Empirical evaluation section] Empirical evaluation section: aggregate statistics are reported across 115,125 negotiations, but the subset of trajectories crossing regime boundaries mid-negotiation is not isolated or analyzed. Without this, the results cannot confirm that the mechanism preserves the claimed regime-specific concession behaviors (quick closure in narrow spreads, savings-matching in wider) under live β recomputation.

    Authors: The referee correctly notes that isolating regime-crossing trajectories would provide stronger confirmation of the adaptive behavior. Our current results aggregate over all 115,125 negotiations (including those with mid-negotiation price revisions that trigger β recomputation) and show overall competitive performance, but we did not break out the crossing subset. In the revised manuscript we will identify these trajectories, report their share of the data, and provide separate metrics for agreement rates, broker savings, and concession speed, comparing them against non-crossing cases and fixed-β baselines. This will directly verify that regime-specific postures are preserved under live β updates. revision: yes

Circularity Check

0 steps flagged

No significant circularity: framework uses explicit deterministic formulas for β and monotonicity with independent empirical validation.

full rationale

The paper defines a spread-derived β and anchor-and-resume mechanism via deterministic formulas that map margin structure to concession posture while enforcing non-decreasing offers. No equation or section reduces the claimed adaptation or monotonicity guarantee to a fitted parameter renamed as prediction, nor to a self-citation chain whose load-bearing premise is unverified. The 115k-negotiation evaluation reports aggregate outcomes against baselines and LLM opponents but does not rely on post-hoc tuning that would make reported performance equivalent to the input construction. The derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The framework depends on the spread-derived beta mapping and the invented anchor-resume mechanism, with limited independent evidence provided in the abstract.

free parameters (1)
  • beta derivation function
    The exact function or parameters for deriving beta from the live spread are not specified, potentially involving choices that affect performance.
axioms (2)
  • domain assumption Offers must be monotonically non-decreasing
    This is stated as a requirement to avoid violating negotiation norms.
  • domain assumption LLM is used only for translation
    Assumed to decouple from pricing logic.
invented entities (1)
  • anchor-and-resume mechanism no independent evidence
    purpose: To guarantee monotonicity under pricing shifts
    Introduced in the paper as a new component without external validation mentioned.

pith-pipeline@v0.9.0 · 5582 in / 1432 out tokens · 35255 ms · 2026-05-09T22:44:21.184025+00:00 · methodology

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

19 extracted references · 19 canonical work pages

  1. [1]

    Jennings

    Peyman Faratin, Carles Sierra, and Nicholas R. Jennings. Nego- tiation decision functions for autonomous agents.Robotics and Autonomous Systems, 24(3–4):159–182, 1998

    work page 1998

  2. [2]

    Cambridge University Press, 2014

    Shaheen Fatima, Sarit Kraus, and Michael Wooldridge.Principles of Automated Negotiation. Cambridge University Press, 2014. 8 Anchor-and-Resume Concession Under Dynamic Pricing for LLM-Augmented Freight Negotiation AAMAS’27, 2027,

    work page 2014

  3. [3]

    Gerding, Koen Hin- driks, Takayuki Ito, Nicholas R

    Tim Baarslag, Katsuhide Fujita, Enrico H. Gerding, Koen Hin- driks, Takayuki Ito, Nicholas R. Jennings, Catholijn Jonker, Sarit Kraus, Raz Lin, Valentin Robu, and Colin R. Williams. Evaluat- ing practical negotiating agents: Results and analysis of the 2011 international competition.Artificial Intelligence, 198:73–103, 2013

    work page 2011

  4. [4]

    Deal or no deal? End-to-end learning for negotiation dialogues

    Mike Lewis, Denis Yarats, Yann Dauphin, Devi Parikh, and Dhruv Batra. Deal or no deal? End-to-end learning for negotiation dialogues. InProceedings of EMNLP, pages 2443–2453, 2017

    work page 2017

  5. [5]

    De- coupling strategy and generation in negotiation dialogues

    He He, Derek Chen, Anusha Balakrishnan, and Percy Liang. De- coupling strategy and generation in negotiation dialogues. In Proceedings of EMNLP, pages 2333–2343, 2018

    work page 2018

  6. [6]

    arXiv preprint

    Yao Fu, Hao Peng, Tushar Khot, and Mirella Lapata. Improving language model negotiation with self-play and in-context learning from AI feedback.arXiv preprint arXiv:2305.10142, 2023

    work page arXiv 2023

  7. [7]

    How well can LLMs ne- gotiate? NegotiationArena platform and analysis

    Federico Bianchi, Patrick John Chia, Mert Yuksekgonul, Jacopo Tagliabue, Dan Jurafsky, and James Zou. How well can LLMs ne- gotiate? NegotiationArena platform and analysis. InProceedings of ICML, 2024

    work page 2024

  8. [8]

    Cooperation, competition, and maliciousness: LLM-stakeholders interactive negotiation

    Sahar Abdelnabi, Amr Gomaa, Sarath Sivaprasad, Lea Schoenherr, and Mario Fritz. Cooperation, competition, and maliciousness: LLM-stakeholders interactive negotiation. InAdvances in Neural Information Processing Systems (NeurIPS), 2024

    work page 2024

  9. [9]

    Kirshner, Yiwen Pan, Jason Xianghua Wu, and Alex Gould

    Samuel N. Kirshner, Yiwen Pan, Jason Xianghua Wu, and Alex Gould. Talking terms: Agent information in LLM supply chain bargaining.Decision Sciences, 57:9–23, 2026

    work page 2026

  10. [10]

    Lundberg and Su-In Lee

    Scott M. Lundberg and Su-In Lee. A unified approach to inter- preting model predictions. InAdvances in Neural Information Processing Systems (NeurIPS), pages 4766–4777, 2017

    work page 2017

  11. [11]

    Automated negotiation for e-commerce decision making: A goal deliberated agent architecture for multi-strategy selection.Deci- sion Support Systems, 73:1–14, 2015

    Mukun Cao, Xudong Luo, Xin Robert Luo, and Xiaopei Dai. Automated negotiation for e-commerce decision making: A goal deliberated agent architecture for multi-strategy selection.Deci- sion Support Systems, 73:1–14, 2015

    work page 2015

  12. [12]

    Opponent modelling in automated multi-issue negotiation using Bayesian learning

    Koen Hindriks and Dmytro Tykhonov. Opponent modelling in automated multi-issue negotiation using Bayesian learning. In Proceedings of the 7th International Joint Conference on Au- tonomous Agents and Multiagent Systems (AAMAS), pages 331–338, 2008

    work page 2008

  13. [13]

    A deep reinforcement learning approach to concurrent bilateral negotiation

    Pallavi Bagga, Nicola Paoletti, Bedour Alrayes, and Kostas Stathis. A deep reinforcement learning approach to concurrent bilateral negotiation. InProceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI), pages 297–303, 2020

    work page 2020

  14. [14]

    PredictRV: A prediction based strategy for negotiations with dynamically changing reservation value

    Aditya Srinivas Gear, Kritika Prakash, Nonidh Singh, and Praveen Paruchuri. PredictRV: A prediction based strategy for negotiations with dynamically changing reservation value. InGroup Decision and Negotiation: A Multidisciplinary Perspective, pages 135–148, Springer, 2020

    work page 2020

  15. [15]

    Nowak and Karl Sigmund

    Martin A. Nowak and Karl Sigmund. Tit for tat in heterogeneous populations.Nature, 355(6357):250–253, 1992

    work page 1992

  16. [16]

    Non- monotonic-offers bargaining protocol.Autonomous Agents and Multi-Agent Systems, 11(1):45–67, 2005

    Pinata Winoto, Gordon McCalla, and Julita Vassileva. Non- monotonic-offers bargaining protocol.Autonomous Agents and Multi-Agent Systems, 11(1):45–67, 2005

    work page 2005

  17. [17]

    Monotonic concession protocols for multilateral ne- gotiation

    Ulle Endriss. Monotonic concession protocols for multilateral ne- gotiation. InProceedings of the 5th International Joint Confer- ence on Autonomous Agents and Multiagent Systems (AAMAS), pages 392–399, 2006

    work page 2006

  18. [18]

    Oil prices soar amid worries of sustained war in Iran.The Washington Post, March 2, 2026

    Evan Halper and Rachel Siegel. Oil prices soar amid worries of sustained war in Iran.The Washington Post, March 2, 2026

    work page 2026

  19. [19]

    $1,450"). - Keep responses concise (2-4 sentences). - If you agree to a rate, say

    Mordor Intelligence. United States freight brokerage market size & share analysis: growth trends and forecast (2026 to 2031). Mordor Intelligence, 2025. https://www.mordorintelligence.com/ industry-reports/united-states-freight-brokerage-market. 9 AAMAS’27, 2027, Hoang Nguyen, Lu Wang, and Marta Gaia Bras A LLM BROKER SYSTEM PROMPT The following is the co...

    work page 2026