arXiv subjects

Jiangwei Pan

Publications and source records attributed to Jiangwei Pan.

3 recordsLinked to original sources

arXiv2026

GenPage: Towards End-to-End Generative Homepage Construction at Netflix

We present GenPage, an end-to-end generative approach to Netflix homepage construction that replaces the traditional multi-stage recommender stack with a single transformer. GenPage treats the user and request context as a prompt and autoregressively generates the entire structured, multi-row homepage as the response. We adapt the LLM training recipe: pretraining on production pages, followed by post-training via weighted binary classification (WBC) or reinforcement learning (RL). For industry-scale deployment, we introduce techniques addressing cold start, model freshness, business-rule enforcement, and serving efficiency. In online A/B tests against a mature, highly optimized production homepage recommender, GenPage delivered a substantial lift on the core user engagement metric we use for launch decisions, while reducing end-to-end serving latency by 20%. Offline, two findings stand out: enriching the prompt yields a larger improvement than scaling model capacity in our current regime, and RL post-training increases homepage diversity even though diversity is not part of the objective.

cs.IR↗

arXiv2016

Approximating Dynamic Time Warping and Edit Distance for a Pair of Point Sequences

We give the first subquadratic-time approximation schemes for dynamic time warping (DTW) and edit distance (ED) of several natural families of point sequences in $\mathbb{R}^d$, for any fixed $d \ge 1$. In particular, our algorithms compute $(1+\varepsilon)$-approximations of DTW and ED in time near-linear for point sequences drawn from k-packed or k-bounded curves, and subquadratic for backbone sequences. Roughly speaking, a curve is $κ$-packed if the length of its intersection with any ball of radius $r$ is at most $κ\cdot r$, and a curve is $κ$-bounded if the sub-curve between two curve points does not go too far from the two points compared to the distance between the two points. In backbone sequences, consecutive points are spaced at approximately equal distances apart, and no two points lie very close together. Recent results suggest that a subquadratic algorithm for DTW or ED is unlikely for an arbitrary pair of point sequences even for $d=1$. Our algorithms work by constructing a small set of rectangular regions that cover the entries of the dynamic programming table commonly used for these distance measures. The weights of entries inside each rectangle are roughly the same, so we are able to use efficient procedures to approximately compute the cheapest paths through these rectangles.

cs.CG↗

arXiv2011

Edit Distance to Monotonicity in Sliding Windows

Given a stream of items each associated with a numerical value, its edit distance to monotonicity is the minimum number of items to remove so that the remaining items are non-decreasing with respect to the numerical value. The space complexity of estimating the edit distance to monotonicity of a data stream is becoming well-understood over the past few years. Motivated by applications on network quality monitoring, we extend the study to estimating the edit distance to monotonicity of a sliding window covering the $w$ most recent items in the stream for any $w \ge 1$. We give a deterministic algorithm which can return an estimate within a factor of $(4+\eps)$ using $O(\frac{1}{\eps^2} \log^2(\eps w))$ space. We also extend the study in two directions. First, we consider a stream where each item is associated with a value from a partial ordered set. We give a randomized $(4+ε)$-approximate algorithm using $O(\frac{1}{ε^2} \log ε^2 w \log w)$ space. Second, we consider an out-of-order stream where each item is associated with a creation time and a numerical value, and items may be out of order with respect to their creation times. The goal is to estimate the edit distance to monotonicity with respect to the numerical value of items arranged in the order of creation times. We show that any randomized constant-approximate algorithm requires linear space.

cs.DS↗