3arXiv cs.CL (Computation and Language)·
STEP: Career-path recommendation system using LLMs, temporal GRU, and contrastive occupation embeddings
Researchers introduce STEP (Sequential Trajectory of Employment Prediction), a career-path recommendation system that extracts structured trajectory data from unstructured resumes using LLMs and models it with a time-decay GRU, FiLM conditioning on educational attainment, and attention-based pooling. They also introduce ROUTE, a two-stage contrastive pretraining procedure for multilingual occupation embeddings. STEP is evaluated on four career-trajectory datasets including an improved version of the public JobHop dataset, outperforming prior baselines on next-job prediction. Code and data are publicly released.
Related guides (1)
Related events (8)
JobHop v2: Large-scale career trajectory dataset extracted from 440K resumes via LLM pipeline
Researchers release JobHop v2, a dataset of 355,315 career trajectories extracted from ~440,000 pseudonymized multilingual resumes provided by VDAB, the Flemish Public Employment Service. The extraction pipeline uses reasoning-controlled LLM inference with a retry mechanism achieving 100% JSON parse rate, annotating trajectories with ESCO occupational codes, temporal information, and education attainment. The work addresses a gap in publicly available, authentic (non-synthesized) career trajectory data for workforce planning and labour market analysis. Dataset and code are publicly released.
STORM: Internalized Spatial-Temporal Reasoning for Video-Language Models via Latent Trajectories
STORMS is a two-stage training framework that teaches large vision-language models to perform spatial-temporal video reasoning through bounded continuous latent trajectories rather than explicit textual chain-of-thought, keyframe selection, or external tool use. In Stage I, latent tokens are aligned with thought-video representations derived from generated videos; in Stage II, answer-only supervision internalizes the reasoning process. At inference time, no video regeneration or frame reinsertion is required, reducing latency and engineering complexity. Evaluations on VideoMME, MVBench, TempCompass, and MMVU show improved accuracy with substantially lower inference overhead versus tool-based pipelines.
TrajTok: Adaptive Spatial Tokenization for Trajectory Representation Learning
TrajTok is a trajectory encoder that learns transferable GPS trace representations via multi-resolution hexagonal spatial tokenization and masked-token pretraining. It uses a factorized transformer with per-modality self-attention, cross-attention fusion, and spatiotemporal rotary position embeddings (ST-RoPE) to jointly encode geometry and kinematics. A single frozen TrajTok encoder with lightweight adapters outperforms task-specific methods on trajectory similarity search, classification, ETA, and travel-time regression on the Porto dataset. The work positions learned spatial tokenization plus masked pretraining as a viable path toward general-purpose trajectory foundation models.
STRACE framework improves LLM agent optimization via structural trajectory analysis and causal extraction
Researchers introduce STRACE (Structural TRajectory Analysis and Causal Extraction), a framework for constructing high signal-to-noise optimization contexts for long-horizon LLM agents. At the batch level, STRACE mines failure patterns to filter redundant traces; within each trace, it performs causal localization over a textual dependency graph to isolate root-cause steps. On the formal verification benchmark VeruSAGE-Bench, STRACE achieves a 1.4× success-rate improvement (42.5% to 58.5%) over human-expert-designed agents, outperforming standard context-filtering baselines.
AgentMob: Training-free LLM agent framework for evidence-grounded mobility prediction
AgentMob is a training-free LLM-driven agent framework that formulates next-location prediction as adaptive evidence-controlled decision making, using a fast path for routine cases and iterative tool use for ambiguous ones. Evaluated on three mobility datasets, it achieves the strongest overall performance among training-free LLM-based methods, with GPT-5.4 reaching 71.42% Acc@1 on the BW dataset. The framework demonstrates that LLM controllers add most value in resolving ambiguous predictions through adaptive evidence gathering rather than routine cases.
DREAM: Training dense retrieval embeddings via autoregressive LLM supervision
DREAM is a new method for training dense retrieval embedding models using the autoregressive next-token prediction objective of a frozen LLM, bypassing the need for labeled positive/negative document pairs required by contrastive training. The approach injects retriever-generated query-document similarity scores into selected attention heads of the LLM, allowing prediction loss gradients to flow back to the retriever. Evaluated on BEIR and RTEB benchmarks with 0.5B–3B parameter backbones, DREAM consistently outperforms contrastive baselines across model scales.
G2Rec: Scalable framework unifying graph-based user modeling with semantic tokenization for generative recommendation
Researchers propose G2Rec, a framework that combines holistic graph-based user co-engagement modeling with semantic tokenization for industrial-scale generative recommendation systems. The approach addresses limitations of existing methods—scalability issues in graph serialization and lack of supervision in semantic tokenization—by learning user interest prototypes without ground-truth labels. The system has been deployed in production across product surfaces and evaluated on public datasets, showing improvements over prior methods.
Progress Advantage: Annotation-Free Step-Level Scoring for LLM Agents via RL Post-Training
Researchers introduce 'progress advantage,' a method that derives implicit step-level reward signals for LLM agents directly from the log-probability ratio between an RL-trained policy and its reference policy, without requiring dedicated process reward model training. The approach is shown to recover the optimal advantage function under a general stochastic MDP formulation, making it annotation-free and domain-agnostic. Validated across five benchmarks and four model families on tasks including test-time scaling, uncertainty quantification, and failure attribution, it outperforms confidence-based baselines and even dedicated trained reward models. The result is practically significant because building process reward models for agentic settings is currently a major bottleneck.