Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Online Bilateral Trade With Minimal Feedback: Don’t Waste Seller’s Time
Authors: Francesco Bacchiocchi, Matteo Castiglioni, Roberto Colomboni, Alberto Marchesi
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we present some experimental results obtained on synthetically-generated instances. Specifically, we consider instances where the seller s valuations are sampled from a Beta distribution with parameters αs and βs, while the buyer s valuations are sampled from a Beta distribution with parameters αb and βb. For each instance, we evaluate the performance of our algorithm and the Scouting Bandits algorithm of Cesa-Bianchi et al. [2021] in terms of cumulative regret. To this end, we run both algorithms on each instance n = 5 times and report the mean and standard deviation of the achieved cumulative regret. Table 1: Comparison between our algorithm and the one of Cesa-Bianchi et al. [2021] in terms of cumulative regret across different instances where buyers and sellers valuations are distributed according to Beta distributions. Table 2: Comparison between our algorithm and the one of Cesa-Bianchi et al. [2021] in terms of cumulative regret across different instances where buyers and sellers valuations are distributed according to Beta distributions. |
| Researcher Affiliation | Academia | Francesco Bacchiocchi Politecnico di Milano EMAIL Matteo Castiglioni Politecnico di Milano EMAIL Roberto Colomboni Politecnico di Milano & Università degli Studi di Milano EMAIL Alberto Marchesi Politecnico di Milano EMAIL |
| Pseudocode | Yes | Algorithm 2 More than one-bit less than two-bit bilateral trade 1: Input: time horizon T 2: Set δ 1 T 2 , H K T 1/3 , Tk,0 0, Qk,0 0, Sk,0 0 k [K] 3: Set PK {pk}k [K] with pk k 1/K k [K], K 4: for t = 1, 2, . . . , T do 5: if t HK then Exploration Phase 6: Set l t/H and post price Pt pl 7: else Bandit Phase 8: Select l argmaxk K UCBk,t 1 (see eq. (3)) and post price Pt pl 9: for k = 1, 2 . . . K do Update Counters 10: if t = HK then 11: Nk mini k Qi,KH 12: K {k [K] | Qk,KH 32 log(KT 2/δ)} 13: ˆFk 1 KH PK 1 i=k PH j=1 I{Bi,j pi}, k K 14: ˆGk 1 KNk Pk i=1 PNk j=1 I{Si,j pi}, k K 15: Set Tk,t Tk,t 1 + I{Pt = pk} 16: Set Qk,t Qk,t 1 + I{Pt = pk}I{Pt Bt} 17: Set Sk,t Sk,t 1 + I{Pt = pk}I{St Pt}I{Pt Bt} 18: Set ˆνk,t Qk,t Tk,t and ˆµk,t Sk,t |
| Open Source Code | No | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [NA] Justification: The paper does not include data or code. |
| Open Datasets | No | In this section, we present some experimental results obtained on synthetically-generated instances. Specifically, we consider instances where the seller s valuations are sampled from a Beta distribution with parameters αs and βs, while the buyer s valuations are sampled from a Beta distribution with parameters αb and βb. ... Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [NA] Justification: The paper does not include data or code. |
| Dataset Splits | No | In this section, we present some experimental results obtained on synthetically-generated instances. Specifically, we consider instances where the seller s valuations are sampled from a Beta distribution with parameters αs and βs, while the buyer s valuations are sampled from a Beta distribution with parameters αb and βb. For each instance, we evaluate the performance of our algorithm and the Scouting Bandits algorithm of Cesa-Bianchi et al. [2021] in terms of cumulative regret. To this end, we run both algorithms on each instance n = 5 times and report the mean and standard deviation of the achieved cumulative regret. Tables 1 and 2 present the comparison. The paper does not provide specific train/test/validation splits as the data is synthetically generated for each run based on Beta distribution parameters, rather than being a fixed dataset that is split. |
| Hardware Specification | No | The paper makes no mention of specific hardware used for running the experiments. The NeurIPS checklist also indicates that the paper does not include experiments (though it does, the detailed info is missing). |
| Software Dependencies | No | The paper does not mention any specific software dependencies or their version numbers that would be needed to replicate the experiment. |
| Experiment Setup | Yes | Algorithm 2 More than one-bit less than two-bit bilateral trade 1: Input: time horizon T 2: Set δ 1 T 2 , H K T 1/3 , Tk,0 0, Qk,0 0, Sk,0 0 k [K] 3: Set PK {pk}k [K] with pk k 1/K k [K], K ... In this section, we present some experimental results obtained on synthetically-generated instances. Specifically, we consider instances where the seller s valuations are sampled from a Beta distribution with parameters αs and βs, while the buyer s valuations are sampled from a Beta distribution with parameters αb and βb. For each instance, we evaluate the performance of our algorithm and the Scouting Bandits algorithm of Cesa-Bianchi et al. [2021] in terms of cumulative regret. To this end, we run both algorithms on each instance n = 5 times and report the mean and standard deviation of the achieved cumulative regret. Table 1 and 2 list parameters such as Time horizon (T), (αs, βs), and (αb, βb). |