On Efficient Recommendations for Online Exchange Markets
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Presently several marketplace applications over online social networks are gaining popularity. An important class of applications is online market exchange of items. Examples include peerflix.com and readitswapit.co.uk. We model this problem as a social network where each user has two associated lists. The item list consists of items the user is willing to give away to other users. The wish list consists of items the user is interested in receiving. A transaction involves a user giving an item to another user. Users are motivated to transact in expectation of realizing their wishes. Wishes may be realized by a pair of users swapping items corresponding to each other's wishes, but more generally by means of users exchanging items through a cycle, where each user gives an item to the next user in the cycle, in accordance with the receiving user's wishes. The problem we consider is how to efficiently generate recommendations for item exchange cycles, for users in a social network. Each cycle has a value which is determined by the number of items exchanged through the cycle. We focus on the problem of generating recommendations under two models. In the deterministic model, the value of a recommendation is the total number of items exchanged through cycles. In the probabilistic model, there is a probability associated with a user transacting with another user and a user being willing to trade an item for another. The value of a recommendation then is the expected number of items exchanged. We show that under both models, the problem of determining an optimal recommendation is NP-complete and develop efficient approximation algorithms for both. We show that our algorithms have guaranteed approximation factors of 2k (for greedy), 2k -1 (for local search), and(2k + 1)/3 (for combination of greedy and local search) where k is the max cycle length. We also develop a so-called maximal algorithm, which does not have an approximation guarantee but is more efficient. We conduct a comprehensive set of experiments. Our experiments show that in practice, the approximation quality achieved by maximal is competitive w.r.t. that of the other algorithms. On the other hand, maximal outperforms all other algorithms on scalability w.r.t. network size.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.002 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it