From Trajectories to Evidence: Auditable Experimental Records for Industrial Research Agents

Research agents increasingly conduct multi-round machine-learning experiments in industrial recommendation settings and retain the resulting trajectories to guide later decisions. Yet a completed trajectory is not automatically evidence: generated artifacts may be unsupported or incomplete, executed rounds may be invalid or confounded, and later modifications may obscure earlier findings. We study \textbf{trajectory-to-evidence conversion}, asking what a completed research process has actually established. We introduce an evidence-grounded framework that couples bounded verification of consequential artifacts with post-execution claim qualification. A context-isolated generate--verify--repair process checks artifacts for evidence violations and missing downstream requirements before release. After execution, validity and attribution checks consolidate evidence across rounds, qualify intervention-level claims as actionable repairs, diagnostic guards, or withheld findings, and preserve admitted claims as auditable records with explicit provenance and applicability boundaries. A hybrid LLM-assisted controller subsequently applies, defers, or rejects records based on available target evidence. Record audits characterize which claims survive qualification, while downstream diagnostics identify affirmative applicability judgment as a bottleneck for the tested controller. Across paper-to-target adaptations, later rounds often improve on the first, while final rounds frequently underperform an earlier best, exposing non-monotonic trajectory evolution. Candidates produced through the complete workflow also yielded positive online lifts relative to deployed baselines.

Reward Guided Decoding for Generative Recommendation

Generative recommendation formulates recommendation task into an SID sequence autoregressive generation paradigm, but the decoding process is often dominated by generation likelihood. This may conflict with real-world business objectives, where high-value candidates can receive low generation probability and be pruned early during beam search. Existing reranking or training-time alignment methods either intervene too late or require costly model retraining when business preferences change. To this end, we propose \textbf{R}eward \textbf{G}uided \textbf{D}ecoding, named \textbf{RGD}, a controllable decoding framework for industrial value-oriented generative recommendation. We formulate value-guided decoding as a KL-regularized reward maximization problem, deriving a closed-form reward guided decoding distribution that principledly combines generation probability with reward signals. RGD treats the base generator as a reference policy and introduces a reward model as a test-time controller, injecting reward into each decoding step to reshape the search trajectory without retraining the generator. Extensive offline and online experiments demonstrate the effectiveness of our approach for aligning personalization and business value. RGD has been deployed on the Kuaishou platform, bringing consistent improvements in real-world recommendation scenarios.

RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation

Multimodal large language models (MLLMs) can convert multimodal item content into structured descriptions used as semantic features for recommendation. Conventional content-only generation, however, cannot use downstream user signals to determine which semantics should be emphasized. Recent user-conditioned methods incorporate these signals through user histories or profiles, but they require user information at inference and make generation user-dependent. In this paper, we introduce RecoReward, which instead uses behavior-derived rewards during training and preserves content-only inference. To instantiate this idea in live-stream recommendation, we treat historically engaged users as a proxy for future target users and use observational non-target users to estimate affinity shared broadly across users. The Recommender Affinity Score (RAS) contrasts these signals to provide user-selective feedback for reinforcement learning, allowing the learned policy to generate a single shared description without user inputs. In our offline benchmark, RecoReward-9B outperforms its Qwen3.5-9B baseline and all other evaluated models across seven recall metrics. Online A/B testing also shows performance gains. These results show that RecoReward trains the MLLM to produce item features that benefit downstream recommendation while retaining content-only serving.

Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence

Modern industrial recommendation systems typically separate recall and ranking into two independent stages. Although this cascade supports corpus-level retrieval and fine-grained multi-objective scoring, it causes objective inconsistency, information loss at the candidate hand-off, and redundant user-side context computation. Meanwhile, the generative recall and ranking scaling share a common Transformer-based modeling philosophy, where architectural consistency creates a natural opportunity for unified integration. However, direct sharing remains challenging since the two tasks require different information visibility and optimization methods. Therefore, we propose \textbf{UniR$^2$}, a \textbf{Uni}fied decoder-only Transformer that unifies Generative \textbf{R}ecall and Multi-Objective \textbf{R}anking within a single heterogeneous sequence comprising user context, SID trajectory, and item features. Within this sequence, the generated trajectory serves as a representation bridge between recall and ranking, where Dual-Query Prefix-Causal Attention provides task-specific visibility. The two tasks share the base attention weights but retain separate optimization boundaries, with ranking-side LoRA preserving ranking adaptability without disrupting the generative backbone. Extensive offline experiments on large-scale industrial data demonstrate the effectiveness and efficiency of UniR$^2$ for both recall and ranking. Long-term online A/B tests on Kuaishou platform further show consistent positive gains, validating the practicality of unified model in large-scale recommendation systems.

Probabilistic Residual Learning for Online Recommendations

Modern recommender systems are typically based on deep learning (DL) models, where a dense encoder learns representations of users and items. As a result, these systems often suffer from the black-box nature and computational complexity of the underlying models, making it difficult to systematically enhance their recommendation capabilities. To address this problem, we propose Probabilistic Residual Learning (PRL), a causal Bayesian recommendation model that models the residual between ground-truth and base predictions, enabling targeted refinement of existing systems. Specifically, PRL (1) probabilistically groups users for localized residual modeling, (2) models domain-level confounders that influence user and item representations, and (3) aggregates cluster-specific residual predictions over the confounders using do-calculus. Experiments demonstrate that our plug-and-play PRL is compatible with various base deep learning recommender systems, improving their performance while automatically discovering meaningful user clusters.

AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems

Recommendation algorithm iteration is moving from an artisanal, engineer-bound process toward an industrialized research loop, but this transition remains blocked by a structural execution bottleneck: the idea-to-launch cycle still depends on human engineers to generate hypotheses, modify production code, launch A/B experiments, and attribute online results. Innovation therefore scales linearly with headcount rather than compounding with evidence, compute, and accumulated experimental knowledge. We present AgentX, a production-deployed multi-agent system that fundamentally restructures this production function. AgentX operates as a self-evolving development engine: it autonomously generates, implements, evaluates, and learns from recommendation experiments at a scale and pace that no manual workflow can sustain. The system orchestrates four tightly coupled stages in a closed loop. A Brainstorm Agent synthesizes evidence from historical experiments, system architecture, data analysis, and external research into ranked, executable proposals. A Developing Agent translates each proposal into production-ready code through repository-grounded generation and multi-dimensional reliability verification. An Evaluation Agent conducts safe online rollout with guardrail-vetoed A/B judgment, converting both successes and failures into structured knowledge assets. A Harness Evolution layer (SGPO) then distills execution trajectories into semantic-gradient updates that continuously sharpen the agents themselves -- making the system not merely automated, but self-improving.

OneReason Technical Report

Generative recommendation models in the OneRec family have been widely deployed in many real-world services, such as short-video, live-streaming, advertising, and e-commerce. However, these generative models can only benefit from the scaling advantage, while their reasoning ability is hard to activate, since we cannot construct meaningful Chain-of-Thought (CoT) sequences consisting of itemic tokens only. Inspired by the success of the reasoning-style ``think before answer'' paradigm in the LLM field, we conduct preliminary studies (i.e., OneRec-Think, OpenOneRec) to explore reasoning capability in generative recommendation. Nevertheless, we notice an unexpected phenomenon: the thinking mode does not show advantages over the non-thinking mode. Drawing insights from recent findings on CoT robustness in multi-modal language models, we argue that effective reasoning in recommendation rests on two factors: perception, the ability to ground itemic tokens in their underlying language semantics, and cognition, the ability to reorganize a user's behavior sequence into coherent latent interest points. We therefore propose OneReason, which includes: (1) strong itemic token perception in pre-training, (2) a three-level cognition-enhanced CoT format for recommendation tasks in SFT, and (3) a specialize-then-unify training recipe in RL to enhance the thinking ability.

ExpWeaver: LLM Agents Learn from Experience via Latent RAG

Experience learning has achieved promising results in enhancing LLM agent planning and reasoning by integrating past interactions as reusable knowledge. However, existing methods remain confined to explicit text space, retrieving experiences via semantic similarity and concatenating them into the context window, leading to substantial token overhead and a decoupled architecture that separates retrieval from generation. To address these limitations, we propose ExpWeaver, a framework that enables LLM agents to learn from experience via latent retrieval-augmented generation, without requiring a separate RAG module. ExpWeaver encodes experiences using the LLM's own hidden states, retrieves relevant experiences directly in latent space at each decoding step, and integrates them through cross-attention aggregation and gated residual mechanisms. The entire pipeline is optimized end-to-end with reinforcement learning, supporting both generative and ranking tasks. We evaluate ExpWeaver on 13 diverse tasks spanning question answering, reasoning, coding, scientific prediction, and recommendation. Results demonstrate that ExpWeaver achieves state-of-the-art performance on 12 out of 13 tasks, outperforming the strongest baseline by over 6.8%; maintains token efficiency comparable to non-retrieval baselines while text-based retrieval methods require 1.5 to 2 times more tokens; and exhibits superior cross-domain generalization, outperforming the strongest baseline by 16.32% under zero-shot transfer and 15.21% under few-shot transfer. Our code for ExpWeaver is released at https://github.com/ulab-uiuc/ExpWeaver.

ExpGraph: Model-Agnostic Experience Learning with Graph-Structured Memory for LLM Agents

Large language model (LLM) agents have shown strong capabilities in reasoning, tool use, and multi-step interaction, but they often solve tasks from scratch and fail to reuse successful strategies or failure lessons from prior experience. Fine-tuning on collected experience can improve reuse, but it is inflexible when stronger or more suitable executors emerge. We propose ExpGraph, a model-agnostic experience learning framework that enables frozen and replaceable LLM executors to improve through external experience reuse without parameter updates. ExpGraph summarizes historical trajectories into reusable skills and failure lessons, organizes them as nodes in a self-evolving experience graph, and retrieves useful experiences through graph diffusion and utility-aware ranking. A lightweight retrieval copilot is trained with reinforcement learning using feedback that compares executor performance with and without retrieved experiences, while the graph is updated online from downstream task outcomes. We evaluate ExpGraph on ExpSuite, covering question answering, mathematical reasoning, code generation, and multi-step agentic environments including ALFWorld and AppWorld. ExpGraph improves over the strongest baseline by 12.2% and 4.7% on static tasks with smaller and larger executors, and by 21.4% and 12.7% in agentic environments, while reducing average interaction steps by 12.7% and 21.6%. Ablations show that graph-structured experience, utility-aware ranking, and adaptive retrieval jointly enable effective experience reuse across diverse tasks and executor models.

LRanker: LLM Ranker for Massive Candidates

Large language models (LLMs) have recently shown strong potential for ranking by capturing semantic relevance and adapting across diverse domains, yet existing methods remain constrained by limited context length and high computational costs, restricting their applicability to real-world scenarios where candidate pools often scale to millions. To address this challenge, we propose LRanker, a framework tailored for large-candidate ranking. LRanker incorporates a candidate aggregation encoder that leverages K-means clustering to explicitly model global candidate information, and a graph-based test-time scaling mechanism that partitions candidates into subsets, generates multiple query embeddings, and integrates them through an ensemble procedure. By aggregating diverse embeddings instead of relying on a single representation, this mechanism enhances robustness and expressiveness, leading to more accurate ranking over massive candidate pools. We evaluate LRanker on seven tasks across three scenarios in RBench with different candidate scales. Experimental results show that LRanker achieves over 30% gains in the RBench-Small scenario, improves by 3-9% in MRR in the RBench-Large scenario, and sustains scalability with 20-30% improvements in the RBench-Ultra scenario with more than 6.8M candidates. Ablation studies further verify the effectiveness of its key components. Together, these findings demonstrate the robustness, scalability, and effectiveness of LRanker for massive-candidate ranking.