GRASPrune: Global Gating for Budgeted Structured Pruning of Large Language Models

Large language models (LLMs) are expensive to serve because model parameters, attention computation, and KV caches impose substantial memory and latency costs. We present GRASPrune, a structured pruning framework applied after pretraining that jointly prunes FFN channels and KV head groups under a single global budget. Instead of learning importance scores without constraints and applying the budget only after training, GRASPrune learns lightweight gate scores with a projected straight-through estimator that enforces a hard mask satisfying the budget at every step while keeping the backbone weights frozen. After the mask is fixed, we calibrate scaling factors on the retained units to mitigate scale mismatch caused by pruning, and fold these factors into the pruned weights to obtain a smaller dense checkpoint with no extra parameters at inference. On LLaMA-2-7B, GRASPrune removes 50% of parameters and achieves 12.18 perplexity on WikiText-2 while maintaining competitive average zero-shot accuracy on five benchmarks, using four epochs on 512 unlabeled calibration sequences on a single NVIDIA A100 80GB GPU without any full model fine-tuning.

Toward Consistent World Models with Multi-Token Prediction and Latent Semantic Enhancement

Whether Large Language Models (LLMs) develop coherent internal world models remains a core debate. While conventional Next-Token Prediction (NTP) focuses on one-step-ahead supervision, Multi-Token Prediction (MTP) has shown promise in learning more structured representations. In this work, we provide a theoretical perspective analyzing the gradient inductive bias of MTP, supported by empirical evidence, showing that MTP promotes the convergence toward internal belief states by inducing representational contractivity via gradient coupling. However, we reveal that standard MTP often suffers from structural hallucinations, where discrete token supervision encourages illegal shortcuts in latent space that violate environmental constraints. To address this, we propose a novel method Latent Semantic Enhancement MTP (LSE-MTP), which anchors predictions to ground-truth hidden state trajectories. Experiments on synthetic graphs and real-world Manhattan Taxi Ride show that LSE-MTP effectively bridges the gap between discrete tokens and continuous state representations, enhancing representation alignment, reducing structural hallucinations, and improving robustness to perturbations.

TabTracer: Monte Carlo Tree Search for Complex Table Reasoning with Large Language Models

Large language models (LLMs) have emerged as powerful tools for natural language table reasoning, where there are two main categories of methods. Prompt-based approaches rely on language-only inference or one-pass program generation without step-level verification. Agent-based approaches use tools in a closed loop, but verification is often local and backtracking is limited, allowing errors to propagate and increasing cost. Moreover, they rely on chain- or beam-style trajectories that are typically combinatorially redundant, leading to high token costs. In this paper, we propose TabTracer, an agentic framework that coordinates multi-step tool calls over intermediate table states, with explicit state tracking for verification and rollback. First, it enforces step-level verification with typed operations and lightweight numeric and format checks to provide reliable rewards and suppress hallucinations. Second, execution-feedback Monte Carlo Tree Search maintains a search tree of candidate table states and uses backpropagated reflection scores to guide UCB1 selection and rollback via versioned snapshots. Third, it reduces redundancy with budget-aware pruning, deduplication, and state hashing with a monotonicity gate to cut token cost. Comprehensive evaluation on TabFact, WikiTQ, and CRT datasets shows that TabTracer outperforms state-of-the-art baselines by up to 6.7% in accuracy while reducing token consumption by 59--84%.

OmniSapiens: A Foundation Model for Social Behavior Processing via Heterogeneity-Aware Relative Policy Optimization

To develop socially intelligent AI, existing approaches typically model human behavioral dimensions (e.g., affective, cognitive, or social attributes) in isolation. Although useful, task-specific modeling often increases training costs and limits generalization across behavioral settings. Recent reasoning RL methods facilitate training a single unified model across multiple behavioral tasks, but do not explicitly address learning across different heterogeneous behavioral data. To address this gap, we introduce Heterogeneity-Aware Relative Policy Optimization (HARPO), an RL method that balances leaning across heterogeneous tasks and samples. This is achieved by modulating advantages to ensure that no single task or sample carries disproportionate influence during policy optimization. Using HARPO, we develop and release Omnisapiens-7B 2.0, a foundation model for social behavior processing. Relative to existing behavioral foundation models, Omnisapiens-7B 2.0 achieves the strongest performance across behavioral tasks, with gains of up to +16.85% and +9.37% on multitask and held-out settings respectively, while producing more explicit and robust reasoning traces. We also validate HARPO against recent RL methods, where it achieves the most consistently strong performance across behavioral tasks.

Affective Flow Language Model for Emotional Support Conversation

Large language models (LLMs) have been widely applied to emotional support conversation (ESC). However, complex multi-turn support remains challenging.This is because existing alignment schemes rely on sparse outcome-level signals, thus offering limited supervision for intermediate strategy decisions. To fill this gap, this paper proposes affective flow language model for emotional support conversation (AFlow), a framework that introduces fine-grained supervision on dialogue prefixes by modeling a continuous affective flow along multi-turn trajectories. AFlow can estimate intermediate utility over searched trajectories and learn preference-consistent strategy transitions. To improve strategy coherence and empathetic response quality, a subpath-level flow-balance objective is presented to propagate preference signals to intermediate states. Experiment results show consistent and significant improvements over competitive baselines in diverse emotional contexts. Remarkably, AFlow with a compact open-source backbone outperforms proprietary LMMs such as GPT-4o and Claude-3.5 on major ESC metrics. Our code is available at https://github.com/chzou25-lgtm/AffectiveFlow.

Entropy-Aware Structural Alignment for Zero-Shot Handwritten Chinese Character Recognition

Zero-shot Handwritten Chinese Character Recognition (HCCR) aims to recognize unseen characters by leveraging radical-based semantic compositions. However, existing approaches often treat characters as flat radical sequences, neglecting the hierarchical topology and the uneven information density of different components. To address these limitations, we propose an Entropy-Aware Structural Alignment Network that bridges the visual-semantic gap through information-theoretic modeling. First, we introduce an Information Entropy Prior to dynamically modulate positional embeddings via multiplicative interaction, acting as a saliency detector that prioritizes discriminative roots over ubiquitous components. Second, we construct a Dual-View Radical Tree to extract multi-granularity structural features, which are integrated via an adaptive Sigmoid-based gating network to encode both global layout and local spatial roles. Finally, a Top-K Semantic Feature Fusion mechanism is devised to augment the decoding process by utilizing the centroid of semantic neighbors, effectively rectifying visual ambiguities through feature-level consensus. Extensive experiments demonstrate that our method establishes new state-of-the-art performance, significantly outperforming existing CLIP-based baselines in the challenging zero-shot setting. Furthermore, the framework exhibits exceptional data efficiency, demonstrating rapid adaptability with minimal support samples.

FlowSteer: Interactive Agentic Workflow Orchestration via End-to-End Reinforcement Learning

In recent years, a variety of powerful agentic workflows have been applied to solve a wide range of human problems. However, existing workflow orchestration still faces key challenges, including high manual cost, reliance on specific operators/large language models (LLMs), and sparse reward signals. To address these challenges, we propose FlowSteer, an end-to-end reinforcement learning framework that takes a lightweight policy model as the agent and an executable canvas environment, automating workflow orchestration through multi-turn interaction. In this process, the policy model analyzes execution states and selects editing actions, while the canvas executes operators and returns feedback for iterative refinement. Moreover, FlowSteer provides a plug-and-play framework that supports diverse operator libraries and interchangeable LLM backends. To effectively train this interaction paradigm, we propose Canvas Workflow Relative Policy Optimization (CWRPO), which introduces diversity-constrained rewards with conditional release to stabilize learning and suppress shortcut behaviors. Experimental results on twelve datasets show that FlowSteer significantly outperforms baselines across various tasks.

Position: Agentic Evolution is the Path to Evolving LLMs

As Large Language Models (LLMs) move from curated training sets into open-ended real-world environments, a fundamental limitation emerges: static training cannot keep pace with continual deployment environment change. Scaling training-time and inference-time compute improves static capability but does not close this train-deploy gap. We argue that addressing this limitation requires a new scaling axis-evolution. Existing deployment-time adaptation methods, whether parametric fine-tuning or heuristic memory accumulation, lack the strategic agency needed to diagnose failures and produce durable improvements. Our position is that agentic evolution represents the inevitable future of LLM adaptation, elevating evolution itself from a fixed pipeline to an autonomous evolver agent. We instantiate this vision in a general framework, A-Evolve, which treats deployment-time improvement as a deliberate, goal-directed optimization process over persistent system state. We further propose the evolution-scaling hypothesis: the capacity for adaptation scales with the compute allocated to evolution, positioning agentic evolution as a scalable path toward sustained, open-ended adaptation in the real world.

Enhancing Language Models for Robust Greenwashing Detection

Sustainability reports are critical for ESG assessment, yet greenwashing and vague claims often undermine their reliability. Existing NLP models lack robustness to these practices, typically relying on surface-level patterns that generalize poorly. We propose a parameter-efficient framework that structures LLM latent spaces by combining contrastive learning with an ordinal ranking objective to capture graded distinctions between concrete actions and ambiguous claims. Our approach incorporates gated feature modulation to filter disclosure noise and utilizes MetaGradNorm to stabilize multi-objective optimization. Experiments in cross-category settings demonstrate superior robustness over standard baselines while revealing a trade-off between representational rigidity and generalization.

LLMdoctor: Token-Level Flow-Guided Preference Optimization for Efficient Test-Time Alignment of Large Language Models

Aligning Large Language Models (LLMs) with human preferences is critical, yet traditional fine-tuning methods are computationally expensive and inflexible. While test-time alignment offers a promising alternative, existing approaches often rely on distorted trajectory-level signals or inefficient sampling, fundamentally capping performance and failing to preserve the generative diversity of the base model. This paper introduces LLMdoctor, a novel framework for efficient test-time alignment that operates via a patient-doctor paradigm. It integrates token-level reward acquisition with token-level flow-guided preference optimization (TFPO) to steer a large, frozen patient LLM with a smaller, specialized doctor model. Unlike conventional methods that rely on trajectory-level rewards, LLMdoctor first extracts fine-grained, token-level preference signals from the patient model's behavioral variations. These signals then guide the training of the doctor model via TFPO, which establishes flow consistency across all subtrajectories, enabling precise token-by-token alignment while inherently preserving generation diversity. Extensive experiments demonstrate that LLMdoctor significantly outperforms existing test-time alignment methods and even surpasses the performance of full fine-tuning approaches like DPO.