arXiv subjects

Chonglin Sun

Publications and source records attributed to Chonglin Sun.

18 recordsLinked to original sources

arXiv2026

GRACE: Generative Recommender Acceleration Engine for Real-Time Ads Retrieval

Productionizing generative recommenders for high-volume, real-time ads retrieval creates two serving challenges: eligibility, ensuring that each generated ad is eligible for the request under the advertiser's audience targeting rules, and compute, which requires meeting strict latency and GPU cost requirements while remaining capable of generating thousands of ads per request with wide-beam decoding. This paper presents GRACE, a serving system for ads generative retrieval that addresses both challenges. For eligibility, GRACE introduces Generative Target Matching (GTM), which extends catalog-valid constrained decoding with personalized filtering over Semantic ID (SID) prefixes using bitmask and Bloom filter matchers derived from targeting rules. SID-level GTM improves final ad-level target matching pass rate from 23.55% to 40.42% over constrained decoding alone. For compute-cost and latency, GRACE targets encoder-decoder Transformers, which are more lightweight than LLMs. It redesigns the decoder around the wide-beam, short-sequence regime, covering attention kernels, KV cache, and beam search optimizations. On NVIDIA GH200, compared with the faster of FlashAttention-2 and FlashAttention-3 baselines, GRACE improves cross-attention latency by 68.0 times and self-attention latency by 23.4-25.8 times across decode steps. Together, these changes reduce decoder latency by 11.1 times, keeping ads generative retrieval within latency and compute requirements.

cs.IR↗

arXiv2026

Self-Guided Test-Time Training for Long-Context LLMs

Long-context processing has become increasingly important for large language models (LLMs), but simply extending the context window does not guarantee effective utilization of long inputs. As input length grows, accuracy often degrades, indicating that models still struggle to identify and use the evidence most relevant to a question. A promising way to improve long-context utilization is test-time training (TTT), which treats the test context as a training example for instance-specific parameter adaptation. However, applying TTT to the entire long context is prohibitively expensive, while adapting on randomly sampled spans introduces severe noise. Because most spans in a long context are irrelevant to the specific question, training on them may even degrade the base model's performance. Our preliminary study shows that TTT is highly sensitive to training-span quality: on LongBench-v2, TTT on randomly sampled spans hurts performance, whereas TTT on oracle spans substantially improves it. Motivated by this, we propose a simple method, Self-Guided TTT (S-TTT): before adaptation, the model identifies the evidence spans it should learn from, and the standard language-modeling training objective is applied only to those selected spans. On two challenging long-context reasoning benchmarks, LongBench-v2 and LongBench-Pro, S-TTT improves accuracy for both Qwen3-4B-Thinking-2507 and Llama-3.1-8B-Instruct, achieving up to a 15% relative improvement.

cs.CL↗

arXiv2026

SCOReD: Student-Aware CoT Optimization for Recommendation Distillation

Chain-of-thought (CoT) distillation in the recommendation domain is a necessary precursor to RL training, but raw teacher traces are ill-suited to this task. Large teachers approach the recommendation task with unusually high reasoning uncertainty, repeatedly rechecking their answers without revising them; supervised fine-tuning on such traces produces verbose students that never revise their initial guess. Furthermore, due to the novelty of the recommendation domain, the teacher's reasoning traces are highly out-of-distribution for the small student LLM. We propose Student-Aware CoT Optimization for Recommendation Distillation (SCOReD), a CoT optimization framework tailored to recommendation that first parses each teacher trace into typed segments and uses the student LLM's attention to score the importance of each segment. Then SCOReD dynamically selects a per-segment edit (KEEP / REWRITE / FUSE / PRUNE) based on the output length and comparative log probability lift of the answer given the edit as per the student. Therefore, SCOReD prunes redundant sections of the reasoning trace while preserving information-dense sections and adapts raw teacher traces to the student's output distribution. Training on SCOReD-optimized CoTs provides a cleaner learning signal to the student model and improves over baseline SFT by 1.56% NDCG and 1.9% Recall@5, while reducing reasoning length by 27.3%.

cs.IR↗

arXiv2026

Diffusion-GR2: Diffusion Generative Reasoning Re-ranker

Generative reasoning re-rankers achieve strong recommendation accuracy by emitting a chain-of-thought before re-ordering a candidate list, but they are slow at inference: an autoregressive (AR) decoder spends one sequential forward pass per reasoning token, and the reasoning trace far exceeds the ranking it produces. To reduce this cost, block-diffusion language models decode many positions in parallel over a few denoising steps and are substantially faster, yet naively converting an AR re-ranker into one opens two accuracy gaps: (1) a structural gap: answer positions are denoised in parallel and scored independently, so the decoder emits invalid rankings (duplicated, dropped, or out-of-set identifiers) that AR avoids through left-to-right masking; and (2) a distributional gap: fine-tuning the converted model on fixed teacher trajectories is off-policy relative to its own decoding at inference, leaving a residual accuracy gap. To close both gaps while keeping the speedup, we propose \textbf{Diffusion-GR2}, a recipe that converts our AR reasoning re-ranker (GR2) into a block-diffusion re-ranker. First, conversion fine-tuning (CFT) adapts the AR-initialized diffusion model to denoise the answer into a valid permutation on its own, without an external constrained decoder. Next, on-policy distillation (OPD) then supervises the model on its own decoded trajectories with dense per-token targets from the AR teacher. Finally, we apply a reinforcement-learning (RL) stage against a re-ranking reward on top of OPD's on-policy policy. Experiments on Amazon Beauty demonstrate that Diffusion-GR2 recovers to near-parity with the AR re-ranker, while block-parallel decoding raises decode throughput by $2.4$--$3.5\times$ at the model's reasoning output length. Ablations show that CFT recovers most of the conversion gap, and that on-policy distillation further closes it to the AR reference.

cs.IR↗

arXiv2026

GR2 Technical Report

Industrial recommendation systems serve billions of users through a multi-stage funnel -- retrieval, early-stage ranking, and re-ranking -- where the final re-ranking step disproportionately shapes user engagement and downstream performance, particularly for carousel and grid display formats. Despite growing enthusiasm for Large Language Models (LLMs) in recommendation, three gaps hinder industrial adoption: (1) most efforts target retrieval and ranking, leaving re-ranking -- the stage closest to the final user experience -- largely underexplored; (2) LLMs are typically deployed zero-shot or via supervised fine-tuning, underutilizing the reasoning capabilities unlocked by reinforcement learning (RL) on verifiable rewards; (3) deployed catalogs index billions of items with non-semantic identifiers that lie outside any base-LLM vocabulary. We present GR2 (Generative Reasoning Re-Ranker), an end-to-end framework that combines (i) mid-training on semantic IDs produced by a tokenizer with >=99% uniqueness, (ii) reasoning-trace distilled from a stronger teacher via targeted prompting and rejection sampling, and (iii) RL with verifiable rewards purpose-built for re-ranking. To make GR2 resource-viable, we further (iv) introduce a context compressor that amortizes training cost, On-Policy Distillation (OPD) as a scalable alternative to SFT -- which we find collapses at industrial scale -- and reasoning distillation for low-latency serving. GR2 delivers +18.7% R@1, +7.1% R@3, and +9.6% N@3 over legacy baselines on industrial-scale traffic. We further find that reward design is critical in re-ranking: LLMs often hack rewards by preserving the incoming order or exploiting position bias, motivating conditional verifiable rewards as essential industrial components.

cs.IR↗

arXiv2026

Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation

Discrete diffusion language models (dLLMs) recover masked tokens in parallel, offering significant speedups over autoregressive (AR) generation. However, such promising frameworks face a fundamental architectural design dilemma: \ding{182} Adopting bidirectional attention achieves strong generation quality by allowing each position to access the full context, but is inherently incompatible with KV caching, limiting inference throughput in batch-serving scenarios; \ding{183} Conversely, causal attention enables efficient cached inference but loses all right-side context, substantially degrading generation quality. This paper introduces Bifocal dLLMs, a new paradigm that resolves this dilemma through \emph{asymmetric bidirectional context}. Analogous to bifocal lenses, we instantiate the paradigm as \textbf{R2LM} (Right-to-Left Mamba), which combines two complementary mechanisms: $a$) standard causal attention providing precise left-context with full KV cache compatibility, while $b$) a lightweight reverse Mamba SSM sidecar supplying compressed right-side context without breaking cacheability. Comprehensive experiments on continued pretraining of Qwen3-1.7B with 60B tokens demonstrate that R2LM achieves $2.4\times$ to $12.9\times$ higher throughput than bidirectional dLLMs and $1.9\times$ to $2.9\times$ speedup over AR baselines in batch serving through parallel decoding with KV caching, while exceeding the causal baseline on most benchmarks and surpassing the bidirectional dLLM on average.

cs.IR↗

arXiv2026

End-to-End Dynamic Sparsity for Resource-Adaptive LLM Inference

Large Language Models (LLMs) inference is typically deployed under a static resource assumption, where models execute a fixed computational graph regardless of the runtime environment. However, real-world cloud infrastructure is inherently dynamic, characterized by fluctuating availability (e.g., spot instance preemption) and tiered Quality-of-Service requirements. In such volatile settings, static models are inflexible: they either crash under resource constraints or waste compute on redundant operations. To bridge this gap, we propose Learning to Allocate (L2A), an end-to-end framework for resource-adaptive inference. Unlike prior methods that condition only on input difficulty, we formulate inference as a constrained allocation problem conditioned on both the input and the runtime resource budget itself. We introduce lightweight, budget-conditioned and input-aware gating networks integrated into the LLM. These gates are trained via a unified objective that jointly optimizes task performance, logical consistency, and resource costs along three axes matching how real-world dynamics manifest: layer skipping for memory and depth pressure, head pruning for throughput contention, and reasoning-token reduction for latency tightening. This lets the model learn a budget-aware policy beyond input difficulty alone: it adaptively configures its computational footprint with respect to real-time resource dynamics, maximizing reasoning depth when resources permit while enforcing strict frugality when budgets tighten. A single L2A model traces the entire compute-accuracy Pareto frontier on Llama-3-8B and Qwen-3-4B: at up to 34% realized layer sparsity, it stays within 0.6% of the dense baseline on GSM8K, with the same gap holding zero-shot on out-of-distribution tasks, while every static or heuristic baseline requires a separately tuned model and still drops by 5-10% at comparable inference time.

cs.IR↗

arXiv2026

Learning Evidence Highlighting for Frozen LLMs

Large Language Models (LLMs) can reason well, yet often miss decisive evidence when it is buried in long, noisy contexts. We introduce HiLight, an Evidence Emphasis framework that decouples evidence selection from reasoning for frozen LLM solvers. HiLight avoids compressing or rewriting the input, which can discard or distort evidence, by training a lightweight Emphasis Actor to insert minimal highlight tags around pivotal spans in the unaltered context. A frozen Solver then performs downstream reasoning on the emphasized input. We cast highlighting as a weakly supervised decision-making problem and optimize the Actor with reinforcement learning using only the Solver's task reward, requiring no evidence labels and no access to or modification of the Solver. Across sequential recommendation and long-context question answering, HiLight consistently improves performance over strong prompt-based and automated prompt-optimization baselines. The learned emphasis policy transfers zero-shot to both smaller and larger unseen Solver families, including an API-based Solver, suggesting that the Actor captures genuine, reusable evidence structure rather than overfitting to a single backbone.

cs.CL↗

arXiv2026

Efficient Retrieval Scaling with Hierarchical Indexing for Large Scale Recommendation

The increase in data volume, computational resources, and model parameters during training has led to the development of numerous large-scale industrial retrieval models for recommendation tasks. However, effectively and efficiently deploying these large-scale foundational retrieval models remains a critical challenge that has not been fully addressed. Common quick-win solutions for deploying these massive models include relying on offline computations (such as cached user dictionaries) or distilling large models into smaller ones. Yet, both approaches fall short of fully leveraging the representational and inference capabilities of foundational models. In this paper, we explore whether it is possible to learn a hierarchical organization over the memory of foundational retrieval models. Such a hierarchical structure would enable more efficient search by reducing retrieval costs while preserving exactness. To achieve this, we propose jointly learning a hierarchical index using cross-attention and residual quantization for large-scale retrieval models. We also present its real-world deployment at Meta, supporting daily advertisement recommendations for billions of Facebook and Instagram users. Interestingly, we discovered that the intermediate nodes in the learned index correspond to a small set of high-quality data. Fine-tuning the model on this set further improves inference performance, and concretize the concept of "test-time training" within the recommendation system domain. We demonstrate these findings using both internal and public datasets with strong baseline comparisons and hope they contribute to the community's efforts in developing the next generation of foundational retrieval models.

cs.IR↗

arXiv2026

SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling

Recent advances in recommendation scaling laws have led to foundation models of unprecedented complexity. While these models offer superior performance, their computational demands make real-time serving impractical, often forcing practitioners to rely on knowledge distillation-compromising serving quality for efficiency. To address this challenge, we present SOLARIS (Speculative Offloading of Latent-bAsed Representation for Inference Scaling), a novel framework inspired by speculative decoding. SOLARIS proactively precomputes user-item interaction embeddings by predicting which user-item pairs are likely to appear in future requests, and asynchronously generating their foundation model representations ahead of time. This approach decouples the costly foundation model inference from the latency-critical serving path, enabling real-time knowledge transfer from models previously considered too expensive for online use. Deployed across Meta's advertising system serving billions of daily requests, SOLARIS achieves 0.67% revenue-driving top-line metrics gain, demonstrating its effectiveness at scale.

cs.LG↗

arXiv2026

GR2: Generative Reasoning Re-ranker

Recent studies increasingly explore Large Language Models (LLMs) as a new paradigm for recommendation systems due to their scalability and world knowledge. However, existing work has three key limitations: (1) most efforts focus on retrieval and ranking, while the reranking phase, critical for refining final recommendations, is largely overlooked; (2) LLMs are typically used in zero-shot or supervised fine-tuning settings, leaving their reasoning abilities, especially those enhanced through reinforcement learning (RL) and high-quality reasoning data, underexploited; (3) items are commonly represented by non-semantic IDs, creating major scalability challenges in industrial systems with billions of identifiers. To address these gaps, we propose the Generative Reasoning Reranker (GR2), an end-to-end framework with a three-stage training pipeline tailored for reranking. First, a pretrained LLM is mid-trained on semantic IDs encoded from non-semantic IDs via a tokenizer achieving $\ge$99% uniqueness. Next, a stronger larger-scale LLM generates high-quality reasoning traces through carefully designed prompting and rejection sampling, which are used for supervised fine-tuning to impart foundational reasoning skills. Finally, we apply Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO), enabling scalable RL supervision with verifiable rewards designed specifically for reranking. Experiments on two real-world datasets demonstrate GR2's effectiveness: it surpasses the state-of-the-art OneRec-Think by 2.4% in Recall@5 and 1.3% in NDCG@5. Ablations confirm that advanced reasoning traces yield substantial gains across metrics. We further find that RL reward design is crucial in reranking: LLMs tend to exploit reward hacking by preserving item order, motivating conditional verifiable rewards to mitigate this behavior and optimize reranking performance.

cs.IR↗

arXiv2025

Haystack Engineering: Context Engineering for Heterogeneous and Agentic Long-Context Evaluation

Modern long-context large language models (LLMs) perform well on synthetic "needle-in-a-haystack" (NIAH) benchmarks, but such tests overlook how noisy contexts arise from biased retrieval and agentic workflows. We argue that haystack engineering is necessary to construct noisy long contexts that faithfully capture key real-world factors -- distraction from heterogeneous biased retrievers and cascading errors in agentic workflows -- to test models' long-context robustness. We instantiate it through HaystackCraft, a new NIAH benchmark built on the full English Wikipedia hyperlink network with multi-hop questions. HaystackCraft evaluates how heterogeneous retrieval strategies (e.g., sparse, dense, hybrid, and graph-based) affect distractor composition, haystack ordering, and downstream LLM performance. HaystackCraft further extends NIAH to dynamic, LLM-dependent settings that simulate agentic operations, where models refine queries, reflect on their past reasonings, and decide when to stop. Experiments with 15 long-context models show that (1) while stronger dense retrievers can introduce more challenging distractors, graph-based reranking simultaneously improves retrieval effectiveness and mitigates more harmful distractors; (2) in agentic tests, even advanced models like Gemini 2.5 Pro and GPT-5 suffer cascading failures from self-generated distractors or struggle to perform early stops. These results highlight persistent challenges in agentic long-context reasoning and establish HaystackCraft as a valuable testbed for future progress.

cs.CL↗

arXiv2025

External Large Foundation Model: How to Efficiently Serve Trillions of Parameters for Online Ads Recommendation

Ads recommendation is a prominent service of online advertising systems and has been actively studied. Recent studies indicate that scaling-up and advanced design of the recommendation model can bring significant performance improvement. However, with a larger model scale, such prior studies have a significantly increasing gap from industry as they often neglect two fundamental challenges in industrial-scale applications. First, training and inference budgets are restricted for the model to be served, exceeding which may incur latency and impair user experience. Second, large-volume data arrive in a streaming mode with data distributions dynamically shifting, as new users/ads join and existing users/ads leave the system. We propose the External Large Foundation Model (ExFM) framework to address the overlooked challenges. Specifically, we develop external distillation and a data augmentation system (DAS) to control the computational cost of training/inference while maintaining high performance. We design the teacher in a way like a foundation model (FM) that can serve multiple students as vertical models (VMs) to amortize its building cost. We propose Auxiliary Head and Student Adapter to mitigate the data distribution gap between FM and VMs caused by the streaming data issue. Comprehensive experiments on internal industrial-scale applications and public datasets demonstrate significant performance gain by ExFM.

cs.IR↗

arXiv2024

Hierarchical Structured Neural Network: Efficient Retrieval Scaling for Large Scale Recommendation

Retrieval, the initial stage of a recommendation system, is tasked with down-selecting items from a pool of tens of millions of candidates to a few thousands. Embedding Based Retrieval (EBR) has been a typical choice for this problem, addressing the computational demands of deep neural networks across vast item corpora. EBR utilizes Two Tower or Siamese Networks to learn representations for users and items, and employ Approximate Nearest Neighbor (ANN) search to efficiently retrieve relevant items. Despite its popularity in industry, EBR faces limitations. The Two Tower architecture, relying on a single dot product interaction, struggles to capture complex data distributions due to limited capability in learning expressive interactions between users and items. Additionally, ANN index building and representation learning for user and item are often separate, leading to inconsistencies exacerbated by representation (e.g. continuous online training) and item drift (e.g. items expired and new items added). In this paper, we introduce the Hierarchical Structured Neural Network (HSNN), an efficient deep neural network model to learn intricate user and item interactions beyond the commonly used dot product in retrieval tasks, achieving sublinear computational costs relative to corpus size. A Modular Neural Network (MoNN) is designed to maintain high expressiveness for interaction learning while ensuring efficiency. A mixture of MoNNs operate on a hierarchical item index to achieve extensive computation sharing, enabling it to scale up to large corpus size. MoNN and the hierarchical index are jointly learnt to continuously adapt to distribution shifts in both user interests and item distributions. HSNN achieves substantial improvement in offline evaluation compared to prevailing methods.

cs.IR↗

arXiv2021

Software-Hardware Co-design for Fast and Scalable Training of Deep Learning Recommendation Models

Deep learning recommendation models (DLRMs) are used across many business-critical services at Facebook and are the single largest AI application in terms of infrastructure demand in its data-centers. In this paper we discuss the SW/HW co-designed solution for high-performance distributed training of large-scale DLRMs. We introduce a high-performance scalable software stack based on PyTorch and pair it with the new evolution of Zion platform, namely ZionEX. We demonstrate the capability to train very large DLRMs with up to 12 Trillion parameters and show that we can attain 40X speedup in terms of time to solution over previous systems. We achieve this by (i) designing the ZionEX platform with dedicated scale-out network, provisioned with high bandwidth, optimal topology and efficient transport (ii) implementing an optimized PyTorch-based training stack supporting both model and data parallelism (iii) developing sharding algorithms capable of hierarchical partitioning of the embedding tables along row, column dimensions and load balancing them across multiple workers; (iv) adding high-performance core operators while retaining flexibility to support optimizers with fully deterministic updates (v) leveraging reduced precision communications, multi-level memory hierarchy (HBM+DDR+SSD) and pipelining. Furthermore, we develop and briefly comment on distributed data ingestion and other supporting services that are required for the robust and efficient end-to-end training in production environments.

cs.DC↗

arXiv2020

Time-based Sequence Model for Personalization and Recommendation Systems

In this paper we develop a novel recommendation model that explicitly incorporates time information. The model relies on an embedding layer and TSL attention-like mechanism with inner products in different vector spaces, that can be thought of as a modification of multi-headed attention. This mechanism allows the model to efficiently treat sequences of user behavior of different length. We study the properties of our state-of-the-art model on statistically designed data set. Also, we show that it outperforms more complex models with longer sequence length on the Taobao User Behavior dataset.

cs.IR↗

arXiv2015

Category Enhanced Word Embedding

Distributed word representations have been demonstrated to be effective in capturing semantic and syntactic regularities. Unsupervised representation learning from large unlabeled corpora can learn similar representations for those words that present similar co-occurrence statistics. Besides local occurrence statistics, global topical information is also important knowledge that may help discriminate a word from another. In this paper, we incorporate category information of documents in the learning of word representations and to learn the proposed models in a document-wise manner. Our models outperform several state-of-the-art models in word analogy and word similarity tasks. Moreover, we evaluate the learned word vectors on sentiment analysis and text classification tasks, which shows the superiority of our learned word vectors. We also learn high-quality category embeddings that reflect topical meanings.

cs.CL↗

arXiv2015

A C-LSTM Neural Network for Text Classification

Neural network models have been demonstrated to be capable of achieving remarkable performance in sentence and document modeling. Convolutional neural network (CNN) and recurrent neural network (RNN) are two mainstream architectures for such modeling tasks, which adopt totally different ways of understanding natural languages. In this work, we combine the strengths of both architectures and propose a novel and unified model called C-LSTM for sentence representation and text classification. C-LSTM utilizes CNN to extract a sequence of higher-level phrase representations, and are fed into a long short-term memory recurrent neural network (LSTM) to obtain the sentence representation. C-LSTM is able to capture both local features of phrases as well as global and temporal sentence semantics. We evaluate the proposed architecture on sentiment classification and question classification tasks. The experimental results show that the C-LSTM outperforms both CNN and LSTM and can achieve excellent performance on these tasks.

cs.CL↗