Paper Daily · All tracks (English)

Pushing Dexterity and Safety for In-Hand Object Pose Reaching That Works Remarkably Well

Thu, 29 Oct 2026 08:00:00 +0800

Skills learned by multi-fingered hands mostly stop at simple grasps or spinning an object in place. For harder in-hand moves the most common failure is dropping the object, yet earlier work rarely counts drops. This paper has a LEAP hand move a held object to almost any position and orientation, with random wrist orientations, through one goal after another. The main policy is trained entirely in…

Embodied AI · Score 4.75 · Ian Chuang, Hiya Shah, Vamsi Chilamkur, Iman Soltani, Claire Tomlin · Source

Exploratory creative pre-ranking: Hierarchical empirical-Bayes Thompson sampling for real-time creative optimization

Thu, 29 Oct 2026 08:00:00 +0800

AI-generated content makes it easy for advertisers to produce many creatives, but each gets very few impressions. This Roblox paper adds a creative pre-ranking layer between recall and ad ranking. The historical play rate μ of a new creative's campaign, game and target audience serves as the prior mean, and the prior strength κ is computed in closed form from the variance of true differences amon…

Personalization · Score 4.75 · Deep dive · 30 min · Jiale Han, Ajith Moparthi, Dan Xie, Xiaowu Dai · Source

EmbeddingGemma 2: Unified and Efficient Multimodal Representations

Thu, 29 Oct 2026 08:00:00 +0800

To search text, images, video and audio with one embedding model, the usual route is to transcribe audio and caption images first, then hand everything to a text model. That costs extra compute and throws away tone of voice and visual detail. Google's EmbeddingGemma 2 is a 744M-parameter model that maps these inputs directly into one vector space, and each modality encoder can be loaded only when…

AI Search · Score 4.65 · Deep dive · 25 min · Henrique Schechter Vera, Sahil Dua, Jovan Andonov, Sara Smoot, Sebastian Russo, Ivan Llanos, Max Ma… · Source

OpenJaws: An Open Hardware UMI Framework with Stereo Localization for Generalist Manipulation

Thu, 29 Oct 2026 08:00:00 +0800

Collecting robot demonstrations in the wild with a handheld gripper (UMI) is now common, but how good the action labels are depends on one step: recovering the gripper trajectory from the wrist-camera video. The original UMI uses a monocular camera plus an IMU (inertial measurement unit); a single camera cannot see true scale, so it also needs a pre-built map and a fiducial marker to stay stable.…

Embodied AI · Score 4.65 · Deep dive · 40 min · Adithyavairavan Murali, Yash Oza, Karanbir Chahal, Neel Jawale, Silas De Vries, Hesam Rabeti, Erwin… · Source

PailitaoMMSearch2.0 Technical Report: Scaling Generative Retrieval to Unified End-to-End Multimodal Search

Thu, 29 Oct 2026 08:00:00 +0800

In Taobao's Pailitao, a user may photograph a shirt and ask for shorts to match it, which calls for a complementary item rather than a duplicate. The usual pipeline classifies intent first and routes to separate retrieval channels, so one wrong step breaks everything after it. Alibaba's report lets a 2B vision-language model read the raw image and instruction and generate product Semantic IDs dir…

Generative · Score 4.50 · Deep dive · 36 min · Xiaohan Ye, Zihan Gong, Jingqian Zhao, Lianyu Du, Yunmeng Shu, Xu Chen, Zesheng Zhou, Renchun You,… · Source

Sequential Asymmetric Imitation for Learning Coupled Robot Policies

Thu, 29 Oct 2026 08:00:00 +0800

When two mobile bimanual robots work together through a shared object, failures usually come not from weak individual skills but from bad timing, such as not waiting or yielding when they should, or still pulling when they should let go. The usual approach needs two people teleoperating at once, which is costly and hard to synchronize. This paper's SAI needs only one operator and works in three s…

Embodied AI · Score 4.40 · Ranpeng Qiu, Yincong Chen, Zihao Li, Yanan Zhou, Guoqiang Ren, Weiming Zhi · Source

DeCAL: Towards Physically-Grounded Dexterous Vision-Language-Action Models via Contact-Aware Latent Co-Imagination

Thu, 29 Oct 2026 08:00:00 +0800

When a dexterous hand does contact-rich manipulation, the fingers often hide the object from the camera, and a vision-only VLA model struggles to handle contact and force. Earlier work has started adding touch, but mostly concatenates tactile and visual features regardless of whether the hand is touching anything. DeCAL lets touch take part according to contact: less when nothing is touched, more…

Embodied AI · Score 4.40 · Yankai Fu, Ning Chen, Junkai Zhao, Heng Zhang, Guocai Yao, Pengwei Wang, Zhongyuan Wang, Shanghang… · Source

A Dense Subset Index for Collective Query Coverage

Thu, 29 Oct 2026 08:00:00 +0800

In classic retrieval, documents compete for the top ranks. In multi-hop QA or text-to-SQL, a single document is often incomplete and the answer must be assembled from several. In dense retrieval terms, the task is to find a small set of documents whose token vectors together cover every token vector of the query. Coverage here means every token vector of the query finds a close token vector somew…

AI Search · Score 4.40 · Kartik Nair, Pritish Chakraborty, Atharva Abhijit Tambat, Indradyumna Roy, Soumen Chakrabarti, Anir… · Source

GRAPE: Let GRPO Supervise Query Rewriting by Ranking for Retrieval

Thu, 29 Oct 2026 08:00:00 +0800

Retrievers like CLIP degrade noticeably on Chinese queries, long-form queries, or combined image-and-text queries, yet retraining the retriever and re-embedding the whole corpus is too expensive. This paper leaves the retriever untouched and trains only an LLM that rewrites the query, using GRPO with retrieval rank as the reward for each rewrite. The authors find that if the reward only looks at…

AI Search · Score 4.40 · Deep dive · 28 min · Zhaohua Zhang, Jianhuan Zhuo, Muxi Chen, Chenchen Zhao, Wenyu Jiang, Mingyang Chen, Yu Tang, Qiuyon… · Source

LTRR: Learning To Rank Retrievers for LLMs

Thu, 29 Oct 2026 08:00:00 +0800

RAG systems usually wire in one fixed retriever, yet different questions suit different retrievers, and some are better answered with no retrieval at all. This paper treats choosing a retriever as a ranking problem. For each training question it records how much each retriever improves answer quality over no retrieval, then trains a ranking model to order the retrievers, with no retrieval as one…

AI Search · Score 4.40 · Deep dive · 29 min · To Eun Kim, Fernando Díaz · Source

VLA-Stem: A Lightweight Safety Mechanism for Vision-Language-Action Models via Reinforcement Learning

Thu, 29 Oct 2026 08:00:00 +0800

VLA models (large models that read images and instructions and output robot actions directly) are good at grasping and placing on command, but most are trained on tables without obstacles; add a box nearby or let something move in, and they drive straight into it. Fine-tuning the VLA on obstacle-avoidance data is the obvious fix, but such data is hard to collect and can hurt the original task ski…

Embodied AI · Score 4.40 · Bojun Zou, Wenhao Yu, Jiaohao Li, Jiaxuan Lin, Guanting Ye, Haohan Geng, Weihao Yu, Yanyong Zhang,… · Source

Rethinking Positional Encoding and Interaction Granularity for CTR Prediction

Thu, 29 Oct 2026 08:00:00 +0800

In unified Transformer CTR models like OneTrans, the authors tried six positional encodings on Kuaishou's KuaiRand data, and none beat using no positional encoding at all. Building on this, the Wuhan University and Tencent paper proposes TRIP (the authors' name for the whole method). Each behavior encodes both its rank in the sequence and how long before this prediction it happened. Behaviors are…

Personalization · Score 4.40 · Deep dive · 30 min · Xinxun Zhang, Yuxiang Wang, Xiaokai Zhou, Jiayue Li, Yi Li, Yue Liu, Jun Zhang, Xiaoling Zhou, Tiey… · Source

Natural Hard-Negative Mining from Positives for Instruction-Following Retrieval

Thu, 29 Oct 2026 08:00:00 +0800

Instruction-following retrieval asks one query to rank documents differently under different instructions: once an instruction adds a requirement, the original positive should drop if it fails that requirement. Training data for this is scarce, and existing methods keep the positive fixed and have an LLM invent a negative passage that violates the instruction. This paper does the reverse. It pick…

AI Search · Score 4.40 · Deep dive · 23 min · Chaolang Li, Jingyu Li, Zhaocheng Du, Xuyang Xie, Gang Wang, Pengwen Dai · Source

TOPS: Reshaping Flow Source and Transport for Multi-Task Robot Manipulation

Thu, 29 Oct 2026 08:00:00 +0800

Multi-task robot policies that generate actions with Flow Matching usually start every task from the same cloud of Gaussian noise and move along a straight path to an action. But opening a drawer and stacking blocks need very different motions, and with one shared start and path it is hard to learn every task well. TOPS keeps the network unchanged and alters only these two things. It traces exper…

Embodied AI · Score 4.40 · Tuo An, Chuhao Zhou, Jiahua Dong, Jingliang Li, Jiaqi Bai, Jindou Jia, Jianfei Yang · Source

CARD: Contact-Anchored Retargeting for Dexterous Manipulation

Thu, 29 Oct 2026 08:00:00 +0800

The easiest way to teach a dexterous robot hand is to convert human hand demonstrations directly into robot actions, which is called retargeting. But existing methods only match fingertip positions and joint angles and ignore whether the hand touches the object and how hard, so the robot hand often fails to stay in contact where it should. CARD first records, for each frame, which parts of the hu…

Embodied AI · Score 4.40 · Cunxi Dai, Botao He, Francois R Hogan, Wei Pan, Rao Fu, Wen-Han Hsieh, Ruoshi Liu, Jitendra Malik,… · Source

Folding Attention Biases into Keys Makes Compacted KV Caches Run on Standard Kernels

Thu, 29 Oct 2026 08:00:00 +0800

KV cache compression can already cut a long document's cache to a tenth, but the strongest method, Attention Matching, adds a bias to every kept key, and vLLM has no slot for it. Using the queries produced while reading the document, this paper computes one direction per attention head on which those queries project to 1 on average, then moves each key along it by the bias times a fixed factor. Q…

Infra · Score 4.40 · Deep dive · 29 min · Utkarsh Kanwat · Source

GRAFT: Geometry-Aware Residual Adaptation for Foundation Trackers

Thu, 29 Oct 2026 08:00:00 +0800

Humanoid behavior foundation models (BFMs, general tracking policies that turn a reference motion into whole-body joint commands) are mostly trained on flat-ground motion and tend to fall on stairs, gaps and platforms. The usual fix is to retrain or fine-tune the whole model together with perception, which is costly and can damage what it already does well on flat ground. Xiaomi Robotics Lab's GR…

Embodied AI · Score 4.40 · Liang Li, Guoyao Zhang, Xingyu Chen, Sikai Wu, Qianyang Wu, Yuhang Lin, Renjing Xu, Mingliang Zhou,… · Source

WAKMan: Workload Aware KVCache Manager

Thu, 29 Oct 2026 08:00:00 +0800

Which eviction policy vLLM's prefix cache should use depends on the traffic arriving right now. Agent requests come back within seconds with the same history, so the default LRU fits. Popular system prompts and long chat histories return much later, and only S3FIFO, which protects blocks used repeatedly, keeps them. Across 20 request mixes, each policy wins some. WAKMan replays the traffic of the…

Infra · Score 4.40 · Deep dive · 26 min · Suhas Kumar Morisetty, Vaibhavi Lokegaonkar, Sachin Mathew Varghese, Wesley Harris · Source

Let the Gate Decide: Per-Lookup Precision for Quantized Conditional Memory in LLMs

Thu, 29 Oct 2026 08:00:00 +0800

When you quantize a language model's n-gram memory tables to low bits, keeping the most frequently accessed rows precise does worse than picking rows at random. The reason is the gate. In conditional memory layers such as DeepSeek's Engram, each fetched vector is scaled by a gate; common function words get small gates, rare entity names get large ones, and quantization error is scaled by the gate…

Infra · Score 4.40 · Deep dive · 35 min · Tianqi Chen, Xunchao Li, Linchuan Xie, Guanghua Yu, Jianchen Zhu, Peisong Wang, Jian Cheng · Source

When Do Discrete Actions Make Good Continuous Controllers?

Thu, 29 Oct 2026 08:00:00 +0800

Large language models predict discrete tokens, yet robot foundation models almost all predict actions as continuous vectors, on the assumption that discretizing loses precision. This paper compares the two systematically, in simulation and on real robots, for single-task policies and multi-task VLAs, and finds that the most naive discrete scheme, cutting each action dimension into evenly spaced b…

Embodied AI · Score 4.40 · Zeji Yi, Soham Dasgupta, Xinyang Zhang, Koushil Sreenath, Guannan Qu, Xi Chen, Max Simchowitz, Guan… · Source

Lossless Compression of Proximity Graph Indexes for Efficient Approximate Nearest Neighbors Search

Thu, 29 Oct 2026 08:00:00 +0800

In HNSW graph indexes for vector search, the vectors are often already quantized, so the edges in the adjacency lists become the main memory cost. Edges are decoded on the fly while each query walks the graph, so compression cannot slow queries down. This work from eBay and the University of Pisa renumbers nodes offline so that neighbors get nearby IDs, then decodes with the SIMD-friendly integer…

Infra · Score 4.30 · Deep dive · 22 min · Francesco Benocci, Leonard Dahlmann, Francesco Gargiulo, Tobias Heuer, Silvio Martinico, Franco Mar… · Source

Tessera: Structure-Completed Evidence Units for Web Retrieval-Augmented Generation

Thu, 29 Oct 2026 08:00:00 +0800

Ask Wikipedia who produced Willy Wonka and the Chocolate Factory, and the answer sits in the producer row of the page's infobox. Text extraction drops the two names in that cell, the reader model sees only the director, and it answers wrong. Tracing answer evidence from page source to reader input over all of English Wikipedia, this paper finds such loss is common: for 7.6% of questions the evide…

AI Search · Score 4.30 · Deep dive · 21 min · Peichun Hua, He Sun, Zirui Zhuang, Qi Qi, Yiming Qiu, Shinan Liu, Yunming Xiao · Source

QESChunker: A Single Objective Unifies Overlapping and Non-Overlapping Chunking for RAG

Thu, 29 Oct 2026 08:00:00 +0800

Most RAG chunking is rule of thumb: fixed length, by paragraph, or by semantic similarity, and none aims directly at whether a question's evidence can be retrieved complete. This paper uses existing questions and their evidence to decide where each document is cut, aiming to put as many evidence passages as possible into chunks that the question can actually retrieve. Without overlap the objectiv…

AI Search · Score 4.30 · Yifan Zhao, Qianyang Zhao, Xinyu Guan, kai wei, Yuming Deng · Source

KL-MOPD: Discrepancy-Aware Multi-Objective On-Policy Distillation for Generative Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

When click and value teachers are distilled together into a generative recommender, equal per-token weights are unfair. On TencentGR the student's KL to the value teacher at the first token is 3.13 times its KL to the click teacher, and the harder teacher switches to the click teacher on the last two tokens. KL-MOPD from Zhejiang University and Alibaba weights each teacher per token by its pre-tr…

Generative · Score 4.20 · Deep dive · 25 min · Weiqin Yang, Yandong Wang, Yuan Gao, Hao Deng, Jinxin Hu, Jing Cao, Jiawei Chen · Source

Latent Residual Reasoning for Ranking in Generative Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

Generative ranking models usually score every candidate in one forward pass. Refiner, from Meituan and the University of Science and Technology of China, adds a few more steps. Each candidate gets a reasoning token, and each step predicts only a correction added to the original score. The base model stays frozen; only LoRA and a new output layer are trained. On Steam it lifts MTGR's AUC from 0.76…

Generative · Score 4.20 · Deep dive · 34 min · Haihan Shi, Ruidong Han, Wenjie Wang, Shangyu Chen, Sunhao Dai, Yongqi Li, Yangyang Li, Fei Jiang,… · Source

Frontier Models for Every Robot: Announcing our API beta

Thu, 29 Oct 2026 08:00:00 +0800

Physical Intelligence has opened its latest robot foundation models as an API beta: developers upload demonstration data, fine-tune a π model on compute it manages, and control their own robots through hosted cloud inference. Early users span many industries, from an autonomous wheel loader filling dump trucks at a real commercial site with no operator on board, to a dual-arm robot deburring part…

Embodied AI · Score 4.10 · Physical Intelligence · Source

REGER: Modeling inter-item Relations with Semantic IDs for Generative Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

Cross-entropy tells a generative recommender only which token is correct, with no signal about how other candidates compare. REGER, from the University of Electronic Science and Technology of China, splits a conventional sequential model's scores over all items into branch preferences along the SID prefix tree and distills them along the true path. Added alone to TIGER, it lifts Beauty Recall@10…

Generative · Score 4.10 · Deep dive · 27 min · Yuxi Huang, Silin Zhou, Chenhao Wang, Zihan Li, Gangming Guo, Pei Ke, Shen Gao, Shuo Shang · Source

Can Complementary Signals Bridge Similarity Islands? Manifold-Augmented Graph Embedding for Multimodal Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

Multimodal recommenders often build item graphs from content similarity, so similar substitutes connect tightly while complements that are bought together but look different (a camera and a memory card) stay isolated. MAGE works in two steps. Offline, it splits each item's co-purchased items by text similarity into substitutes and complements. The text vectors of the item and its substitutes form…

Personalization · Score 4.10 · Zhenye Yang, Jinpeng Chen, Wenbo Fu, Jun Ma, Huan Li, Qimin Zhou, Tao Wang, Li Sun, Senzhang Wang · Source

LESSRec: Level-Specific History Modeling for Semantic ID-Based Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

When a semantic-ID recommender predicts a token at one level, it attends to historical tokens at every level. LESSRec's inference-time analysis finds that history at the target token's own level is more informative and other levels can interfere. So it encodes history separately per level, attends only to the matching level during prediction, and adds an item-level semantic sequence, fusing both…

Generative · Score 4.10 · Shangxin Yang, Zongwei Wang, Jing Long, Junliang Yu, Min Gao · Source

TAO: Tool-Aware Joint Scheduling for LLM-Based Agentic Serving

Thu, 29 Oct 2026 08:00:00 +0800

In agent serving, model inference runs on GPU clusters while tools run on distributed CPUs, and existing schedulers ignore when tool results will be ready and use heuristics to decide which sessions get a GPU and where. For each session and node, TAO computes three values. The first is how much cache it can reuse there, decaying over time. The second is whether its tool will return within the nex…

Infra · Score 4.10 · Shuhuai Li, Shengdong Ye, Dongdong Ge, Yinyu Ye, Jianghao Lin · Source

Compote: Collapse-Resistant Multiple Hypothesis Learning For Sequential Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

A common problem in multi-interest retrieval is that the interest vectors end up pointing the same way. Criteo and Sorbonne's COMPOTE (the authors' name, short for collapse-resistant multiple-hypothesis learning) puts several prediction heads on a shared causal Transformer. Each sample splits its gradient across heads in proportion to the probability each gives the true item, raised to the power…

Personalization · Score 4.10 · Deep dive · 28 min · Louis Castres Saint Martin, Clement Calauzenes, Alain Rakotomamonjy, Patrick Gallinari · Source

JevRec: Linguistic Encoding with Jev-like Decision Modeling for LLM-based Generative Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

Most generative recommenders give every item an SID of the same length, so popular items keep paying for fine-grained codes, while shorter codes squeeze long-tail semantics. JevRec borrows the natural-language pattern that frequent words are short: popular items get short SIDs and long-tail items long ones. A dedicated head predicts the length first and a separate head generates content condition…

Generative · Score 4.10 · Deep dive · 30 min · Qiyong Zhong, Jiajie Su, Ming Yang, Li Zhang · Source

BehaveRec: When Does Long-Term Behavioral History Help Sequential Recommendation? A Study with Session, History, and Community Signals

Thu, 29 Oct 2026 08:00:00 +0800

Are long-term history and community signals always useful? BehaveRec's answer is that it depends on how much history a user has. It combines three signals: a causal Transformer over the current session, a bidirectional encoder over 90 days of history, and a user graph built from behavioral similarity, with a gate deciding per user how much of each to use. On industrial data with over ten million…

Personalization · Score 4.10 · Aysenur Inan, Reza Yousefi Maragheh, Jayesh Uddhav Kudase, Selene Xu, Kaushiki Nag, Jianpeng Xu, Su… · Source

RELAY: One Query, Two Roles for Sequential Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

In sequential recommendation, the part that represents user preference and the history used for scoring often look at different evidence. RELAY has one learned query attend over the history to form a preference representation, then reuses the same attention weights for two memories. Direct history boosts items the user already engaged with. Transition memory holds training-set counts of item b fo…

Personalization · Score 4.10 · Zesheng Li, Chengchang Pan, Honggang Qi · Source

ResCom: Activation Residual Compensation for Large Language Model Quantization

Thu, 29 Oct 2026 08:00:00 +0800

When both weights and activations are quantized to low bits, activations in the same quantization bin collapse to one value and accuracy drops. ResCom takes back the discarded rounding residual to compensate the layer-output error. On calibration data it measures the residual's covariance and the cross-covariance between output error and residual. It uses the first to whiten the residual's scales…

Infra · Score 4.10 · Ning Yang, Bohan Cui, Zhuowen Liu, Yuting Wang · Source

Lost in the bf16 Cast: Exporting Ternary Language Models Can Revert Most Low-Learning-Rate Code Changes

Thu, 29 Oct 2026 08:00:00 +0800

When ternary language models are released, the export first casts latent weights to bf16, and this step quietly erases most of what low-learning-rate fine-tuning changed. Ternary models are fine-tuned through higher-precision latent weights, and a ternary code changes only when its latent crosses a threshold at half the mean magnitude of its tensor. At low learning rates, most changed latents end…

Infra · Score 4.10 · Avichal Sahai, Nishant Raj, Animesh Srivastava · Source

TaskBridge: Technical Report for SIGIR 2026 eCommerce Challenge

Thu, 29 Oct 2026 08:00:00 +0800

Many e-commerce queries describe a task rather than a product. A user types make lasagna and actually needs a baking dish, ingredients and kitchenware, but product titles contain none of those words, so BM25 struggles to retrieve them. This winning solution to an e-commerce challenge first aligns the query with category and entity terms in the catalog, then has an LLM generate five complementary…

AI Search · Score 4.05 · Zhiheng Fu, Yupeng Hu, Zixu Li, Zhiwei Chen, Qinlei Huang, Shaokun Wang, Han Liu, Fan Liu, Weili Gu… · Source

Cross-Day Compatibility of Daily Item Embeddings: An Empirical Study of Alignment, Scale Management, and CTR Adaptation

Thu, 29 Oct 2026 08:00:00 +0800

Many teams retrain item co-occurrence embeddings daily and feed them to CTR models, but the same item's vector changes direction and length every day, so multi-day training data actually mixes several versions. This industrial empirical study compares many ways to handle it and reaches three conclusions. First, orthogonally aligning each new day's vectors to the previous day makes vectors more co…

Personalization · Score 4.05 · Yicheng Wang · Source

Memory-Enforced GPU Sharing in Kubernetes with NVIDIA CUDA 13.4

Thu, 29 Oct 2026 08:00:00 +0800

Many AI workloads need only a fraction of a GPU, yet existing sharing modes fall short: time-slicing does not limit memory, and MPS, which runs processes concurrently, has no application-transparent memory cap. This NVIDIA blog post introduces the memory boundary added in CUDA 13.4: a workload gets exactly the memory it asks for and allocations beyond it fail, or it gets a guaranteed minimum with…

Infra · Score 4.00 · Source

Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design

Thu, 29 Oct 2026 08:00:00 +0800

Meta's next-generation ads retrieval engine Andromeda brings near-ranking-quality models to retrieval, scoring every eligible ad. The new version replaces the single user tower with several specialized user towers as experts, and encodes the ad tower with real-time request context. An attention-based gate in the interaction layer activates only the most relevant experts per ad, cutting inference…

Personalization · Score 4.00 · Deep dive · 30 min · Source

DuTri: Dual-Branch Modeling with Triadic Factorization for Multimodal Multi-Scenario Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

When multi-scenario recommenders add text and image information, they usually feed the vectors as ordinary features into the original parameter-sharing structure. DuTri's authors show with controlled experiments that this is not optimal, so they split the model in two. The collaborative branch applies CP decomposition to the three-way tensor of user, scenario and item interactions, yielding user,…

Personalization · Score 4.00 · Jintao Huang, Chaohua Yang, Kailiang Hao, Dugang Liu, Xing Tang, Zhong Ming · Source

RouteDRAG: Learning to Route and Control Agentic Execution for Multimodal Document RAG

Thu, 29 Oct 2026 08:00:00 +0800

In multimodal QA over long documents, agent-style repeated retrieval and reasoning is not always useful: some questions are answered right after one retrieval, and extra steps can turn a right answer wrong. The authors first show experimentally that one more tool call can improve, preserve or degrade the answer. RouteDRAG therefore learns two things: whether a question should go to the agent at a…

AI Search · Score 4.00 · Zekai Chen, Hao Zhu, Yu Liu, Xunkai Li, Zejian Kang, Rui Qian, Quanjiang Li, Rong-Hua Li · Source

LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets

Thu, 29 Oct 2026 08:00:00 +0800

When one base model serves many LoRA adapters, compressing them under a shared rank budget means deciding how many ranks each layer of an adapter keeps and how the budget is split across adapters. FRA needs no labels: on a small set of unlabeled inputs it compares each adapter's output distribution before and after compression using JS divergence. It measures this for truncating each module, buil…

Infra · Score 4.00 · Shenghui Li, Thiemo voigt · Source

When Connectivity Misleads: Guiding Agentic Search with Contrastive Evidence Paths

Thu, 29 Oct 2026 08:00:00 +0800

When a search agent looks for evidence on a document graph, candidates are often scored with Personalized PageRank. Well-connected documents are naturally easy to reach, so their scores are inflated even when their evidence is no more useful, and they crowd the agent's limited observation window. This paper scores each candidate by one strongest path only, the one from the initially retrieved doc…

AI Search · Score 4.00 · Junchen Li, Liang Xu, Rongzheng Wang, Qizhi Chen, Junhao Wen, Shuang Liang · Source

Layerwise Repacking: Variable-Length Hybrid-Attention Training without an Extra Redistribution Pass

Thu, 29 Oct 2026 08:00:00 +0800

When training hybrid-attention models on variable-length data, the longest samples need wide sequence-parallel groups spanning dozens of GPUs in full-attention layers, while sparse-attention layers balance within single-node groups. Reusing the wide groups pushes sparse-layer communication across nodes, and switching groups separately would add another data transfer. Full-attention layers already…

Infra · Score 4.00 · Jiahao Zhang, Junwei Feng, Dongliang Huang, Xiaoling Zhou, Yiqiao Liao, Chengru Song · Source

Modeling Preference Dependencies with Directed Acyclic Graphs for Narrative-Driven Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

When users describe needs in free text, preferences often qualify one another, as in a seaside city that suits elderly parents but is not too far away. Existing methods treat the text as flat keywords or hand it to an LLM to generate queries freely. PDG-F first has an agent parse the request into a directed acyclic preference dependency graph, then retrieves candidates from the open web along gra…

Generative · Score 3.85 · Mingzhe Lu, Jian Qin, Ruize Qin, Jinhao Liu, Yunpeng Li, Hang Su · Source

Preserving rare-item recall under a storage budget: quantization, distillation, and the limits of teacher fidelity

Thu, 29 Oct 2026 08:00:00 +0800

With a fixed embedding storage budget, should you use a small full-precision model, a distilled student, or a higher-dimensional low-precision model? This paper compares several options with exactly the same byte count on six datasets, focusing on recall of rare items. The result is consistent. Factorizing a large teacher's user-by-item score matrix into fairly high-dimensional user and item vect…

Personalization · Score 3.85 · Han Chen, Yingrui Li · Source

RPSD: Retrieval-Privileged On-Policy Self-Distillation with Optimal Transport

Thu, 29 Oct 2026 08:00:00 +0800

When an LLM recommender is post-trained with GRPO, one sequence-level reward is shared by every token, and each signal comes from a single user. RPSD has a teacher with privileged information re-evaluate the student's own response, and uses their per-token confidence gap to turn the sequence reward into token-level credit. The privileged information comes from similar users: time-aware optimal tr…

Generative · Score 3.75 · Dongyue Li, Zebang Cheng, Zhongyi Deng, Hongboxu, Yi Bin, Fei Ma · Source

The Vocabulary Is Already Here: Adapting Pretrained Interpretable Features for Generative Recommendation

Thu, 29 Oct 2026 08:00:00 +0800

Existing generative recommenders train a dedicated SID vocabulary fitted to language-model representations of item descriptions. This paper notes that pretrained sparse autoencoders already expose language-model features with human-readable annotations, so feature indices can serve directly as the SID vocabulary. The API framework keeps each item's selected feature indices and relative activation…

Generative · Score 3.75 · Ziluowen Luo, Ruochen Liu, Jun Yin, Hao Yan, Zhengxuan Hu, Xinze Shi, Chaozhuo Li, Chengqi Zhang, S… · Source

MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations

Thu, 29 Oct 2026 08:00:00 +0800

Judging recommendation explanations by hand is accurate but expensive, while a single LLM judge is prompt-sensitive and inconsistent. MAREval splits the job across agents: a planner assigns roles with debate-style prompting, a moderator keeps the discussion on track, an arbitrator aggregates multiple rounds, and Monte Carlo sampling adds robustness. On a public benchmark and an e-commerce dataset…

Generative · Score 3.65 · Reza Yousefi Maragheh, Jayesh Uddhav Kudase, Aysenur Inan, Ramin Giahi, Kai Zhao, Parth Parikh, Jia… · Source