Papers by Ke Zeng

25 papers
VANE: Guiding High-Value Exploration in RLVR via Outcome-Process Novelty Shaping (2026.findings-acl)

Copied to clipboard

Challenge: Extensive experiments on large-scale mathematical reasoning and out-of-distribution tasks demonstrate the effectiveness and generalization of the proposed method.
Approach: They propose a method that quantifies novelty across the outcome space and semantic process space by using reward or solution divergence.
Outcome: Experiments on Qwen2.5-Math-7B demonstrate the proposed method is general and efficient.
SILO-BENCH: A Scalable Environment for Evaluating Distributed Coordination in Multi-Agent LLM Systems (2026.acl-long)

Copied to clipboard

Challenge: Existing benchmarks conflate coordination ability with role-based priors.
Approach: They propose a role-free benchmark for evaluating free-form collaboration under information silos.
Outcome: The proposed benchmark systematically probes coordination capabilities under information silos using 54 configurations and 3 frontier LLMs.
Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects (2026.findings-acl)

Copied to clipboard

Challenge: Large Vision-Language Models are hindered by a systemic efficiency barrier known as visual token dominance.
Approach: They propose a systematic taxonomy of efficiency techniques structured around the inference lifecycle . they examine visual encoding, prefilling, and decoding to understand bottlenecks .
Outcome: The proposed techniques reveal how upstream decisions dictate downstream bottlenecks . the proposed techniques include hybrid compression and modality-aware decoding .
Don’t Half-listen: Capturing Key-part Information in Continual Instruction Tuning (2025.acl-long)

Copied to clipboard

Challenge: Existing methods to improve instruction tuning for large language models may cause catastrophic forgetting (CF) CF is a problem where previously learned abilities are degraded .
Approach: They propose a continual instruction tuning method that uses key-part information gain to replay data and refine training objective.
Outcome: The proposed method achieves superior performance on both seen and held-out tasks.
Learning or Self-aligning? Rethinking Instruction Fine-tuning (2024.acl-long)

Copied to clipboard

Challenge: Instruction fine-tuning (IFT) is a crucial phase in building large language models (LLMs).
Approach: They propose a knowledge intervention framework to decouple the potential underlying factors of IFT and enable individual analysis of different factors.
Outcome: The proposed framework decouples the potential underlying factors of IFT, enabling individual analysis of different factors.
Dual-Stage Multi-Task Syntax-Oriented Pre-Training for Syntactically Controlled Paraphrase Generation (2024.findings-acl)

Copied to clipboard

Challenge: Syntactically controlled paraphrase generation (SCPG) aims to generate sentences with syntactic structures resembling given exemplars.
Approach: They propose a dual-stage multi-task pre-training scheme that uses a series of structure-oriented and syntax-oriented tasks to generate sentences with syntactic structures resembling given exemplars.
Outcome: The proposed method outperforms existing methods on all possible variants of SCPG tasks and significantly outperformed the popular T5 model.
Turning Failures into Value: Negative Experience Replay for RLVR via Confidence Gating and Boundary Failure Sampling (2026.acl-long)

Copied to clipboard

Challenge: Existing experience replay methods for RLVR ignore sample inefficiency . expensive reasoning trajectories are discarded immediately after a single gradient update .
Approach: They propose a method to replay failure trajectories to improve model refinement . they propose 'nexGRPO' which employs mid-confidence gating to filter invalid noise and saturated errors.
Outcome: The proposed model outperforms strong baaselines and achieves improved out-of-distribution generalization.
MTP-RL: Acceleration of Reinforcement Learning Rollouts with Policy-Aligned Multi-Token Prediction (2026.findings-acl)

Copied to clipboard

Challenge: Reinforcement learning (RL) is widely applied to boost the performance of pretrained models, yet its training efficiency is severely constrained by rollout generation.
Approach: They propose a framework that accelerates the rollout phase for diverse models by equipping a pipeline to equip the multi-layer parameter-sharing MTP for all models and an advantage-aware MTP optimization strategy.
Outcome: The proposed framework achieves stable growth of acceptance length during RL training, and also accelerates RL rollouts, achieving an average 23.1%–55.3% reduction in rollout time compared to baselines.
DIFFA-2: A Practical Diffusion Large Language Model for General Audio Understanding (2026.findings-acl)

Copied to clipboard

Challenge: Autoregressive (AR) large audio language models are expensive in data and computation . prior work shows diffusion-based LALMs can improve audio understanding under matched settings .
Approach: They propose a diffusion-based LALM that upgrades the speech encoder and employs dual semantic and acoustic adapters.
Outcome: a new model improves over existing autoregressive large language models and is competitive to strong AR models . the proposed model can make use of limited training data and improve inference efficiency . a recent study shows that diffusion-based models can improve audio understanding .
A Reasoner for Real-World Event Detection: Scaling Reinforcement Learning via Adaptive Perplexity-Aware Sampling Strategy (2025.emnlp-industry)

Copied to clipboard

Challenge: Existing methods for abnormal event detection face two predominant limitations . existing methods rely on specialized small models and are limited by performance bottlenecks .
Approach: They propose a framework that leverages the advanced reasoning capabilities of large language models for abnormal event detection.
Outcome: The proposed framework achieves the highest F1 score and an average improvement of 9.59% in OOD transfer tests.
HybridKV: Hybrid KV Cache Compression for Efficient Multimodal Large Language Model Inference (2026.acl-long)

Copied to clipboard

Challenge: Multimodal Large Language Models (MLLMs) are hindered by the rapid growth of key–value (KV) caches.
Approach: They propose a hybrid KV cache compression framework that reduces KV memory by 7.9 and speeds up decoding by 1.52.
Outcome: Experiments on 11 multimodal benchmarks show that HYBRIDKV cuts KV cache memory by 7.9 and speeds up decoding by 1.52.
PaTaRM: Bridging Pairwise and Pointwise Signals via Preference-Aware Task-Adaptive Reward Modeling (2026.acl-long)

Copied to clipboard

Challenge: Existing reward models lack generative and reasoning capabilities, resulting in poor performance.
Approach: They propose a reward-aware task-adaptive reward model that enables pointwise training using readily available pairwise data via a novel Preference-Aware Reward mechanism.
Outcome: The proposed reward model achieves an average relative improvement of 8.7% over the base models on RewardBench and RMBench.
MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning (2026.acl-long)

Copied to clipboard

Challenge: Existing RLVR algorithms rely on rigid, uniform, and symmetric trust region mechanisms . current algorithms lack robustness, asymmetric signal reliability and inefficient gradient utilization .
Approach: They propose a framework to harmonize three dimensions of RLVR algorithms, a paper argues . a binary cutoff is used to discard valuable reinforcement signals, they argue .
Outcome: The proposed framework outperforms baselines in evaluating a robust RLVR solution.
From log 𝜋 to 𝜋: Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight (2026.acl-long)

Copied to clipboard

Challenge: Standard algorithms for Large Language Models (LLMs) enforce stability via "hard clipping" but relying on log-probability gradient yields divergent weights as probabilities vanish, destabilizing LLM training.
Approach: They propose a decoupled gradient policy optimization that uses a decay mechanism to decouple the probability of a boundary token.
Outcome: The proposed algorithm outperforms baselines on various mathematical benchmarks.
IF-CRITIC: Towards a Fine-Grained LLM Critic for Instruction-Following Evaluation (2026.acl-long)

Copied to clipboard

Challenge: Existing evaluation models for instruction-following have many shortcomings, such as substantial costs and unreliable assessments.
Approach: They propose an LLM critic for fine-grained instruction-following evaluation using a checklist generator and a constraint-level preference optimization method.
Outcome: The proposed model beats strong LLM-as-a-Judge baselines in evaluations under lower computational overhead compared to baselines.
When to Continue Thinking: Adaptive Thinking Mode Switching for Efficient Reasoning (2025.findings-emnlp)

Copied to clipboard

Challenge: Large reasoning models (LRMs) incur excessive computational overhead due to redundant reasoning, especially on simple tasks.
Approach: They propose an Adaptive Self-Recovery Reasoning framework that suppresses unnecessary reasoning and enables implicit recovery.
Outcome: The proposed framework suppresses unnecessary reasoning and enables implicit recovery.
Enhancing Efficiency and Exploration in Reinforcement Learning for LLMs (2025.emnlp-main)

Copied to clipboard

Challenge: Existing approaches allocate an equal number of rollouts to all questions during the RL process, which is inefficient.
Approach: They propose a mechanism for dynamically allocating rollout budgets based on the difficulty of the problems, enabling more efficient RL training.
Outcome: The proposed model improves response precision while preserving exploratory ability to uncover potential correct pathways.
Fin-STAR: Structure-as-Semantics to Resolve Implicitness in Financial Retrieval (2026.findings-acl)

Copied to clipboard

Challenge: Existing Retrieval-Augmented Generation systems treat structure as a physical navigational skeleton rather than intrinsic semantic knowledge.
Approach: They propose a framework that redefining hierarchy as intrinsic semantics and uses snippets to enrich hierarchical lineage.
Outcome: The proposed framework outperforms state-of-the-art hierarchical and graph-based benchmarks on FinTierQA Gold.
Global Context or Local Detail? Adaptive Visual Grounding for Hallucination Mitigation (2026.findings-acl)

Copied to clipboard

Challenge: Large vision–language models suffer from object-existence hallucinations when multi-step deliberation decouples from visual evidence.
Approach: They propose a framework that allocates visual computation by uncertainty . they propose highlighting retains global context, while selective zoom-in performs local verification.
Outcome: The proposed framework reduces the complexity of multimodal reasoning by minimizing the operator trade-off.
When 20 Agents Fail to Sort: The Distributed Sorting Benchmark for Scalable Multi-Agent Systems (2026.findings-acl)

Copied to clipboard

Challenge: MAS-BENCH isolates coordination under explicit communication constraints . CAMOC significantly improves coordination success and efficiency across backends .
Approach: They propose a distributed-sorting benchmark that isolates coordination under explicit communication constraints.
Outcome: MAS-BENCH improves coordination success and efficiency across backends . CAMOC significantly improves efficiency under shared-state interaction .
How to Allocate, How to Learn? Dynamic Rollout Allocation and Advantage Modulation for Policy Optimization (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods for reinforcement learning with verifiable rewards are limited by the complexity of the problem and the complexity.
Approach: They propose a theoretically-grounded dual-pronged optimization framework for reinforcement learning with verifiable rewards that compensates for gradient attenuation of high-confidence correct actions while utilizing entropy changes as computable indicators to stabilize excessive update magnitudes.
Outcome: The proposed framework compensates for gradient attenuation of high-confidence correct actions while utilizing entropy changes as computable indicators to stabilize excessive update magnitudes.
CritiqueLLM: Towards an Informative Critique Generation Model for Evaluation of Large Language Model Generation (2024.acl-long)

Copied to clipboard

Challenge: Existing models for NLP evaluations lack the ability to generate informative critiques in pointwise grading and pairwise comparison especially without references.
Approach: They propose a method which can acquire pointwise grading critiques with pseudo references and revise these critiques via multi-path prompting to obtain informative evaluation data in different tasks and settings.
Outcome: The proposed method outperforms all open-source models and even GPT-4 in system-level correlations of pointwise grading.
OssCSE: Overcoming Surface Structure Bias in Contrastive Learning for Unsupervised Sentence Embedding (2023.emnlp-main)

Copied to clipboard

Challenge: Recent studies show that contrastive learning is effective in sentence representation learning . but, the surface structure bias is a problem in the current model .
Approach: They propose to combine a sentence with a sub-semantic sentence to investigate the surface structure bias.
Outcome: The proposed model achieves state-of-the-art on standard semantic textual similarity tasks using different pre-trained backbones.
Harmonizing Dense and Sparse Signals in Multi-turn RL: Dual-Horizon Credit Assignment for Industrial Sales Agents (2026.acl-industry)

Copied to clipboard

Challenge: Large language models for industrial sales require balancing long-term commercial objectives with immediate linguistic constraints such as fluency and compliance.
Approach: They propose a framework that disentangles optimization across time scales by normalizing advantages from turn-level and session-level rewards before fusion.
Outcome: The proposed framework outperforms the state-of-the-art GRPO model in conversion rate and identity detection rate.
CMNEE:A Large-Scale Document-Level Event Extraction Dataset Based on Open-Source Chinese Military News (2024.lrec-main)

Copied to clipboard

Challenge: Current research focuses on the general news or financial domains, with relatively few studies for military domain.
Approach: They propose to annotate Chinese military news events from documents using a schema for the military domain.
Outcome: The proposed dataset is large-scale, document-level open-source for the military domain . it contains 17,000 documents and 29,223 events, which are all manually annotated .

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations