2 citations · 2 across the 13 of their papers we have counts for
14 papers
NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness
NeoHorse Team, Guoliang Cao, Guohao Dai +34
Recursive self-improvement (RSI) requires a concrete mechanism through which an AI system observes its capabilities and converts that evidence into the next round of learning. We p…
DiscoLoop: Looping Discrete Embeddings and Continuous Hidden States for Multi-hop Reasoning
Hengyu Fu, Tianyu Guo, Zixuan Wang +5
Large language models achieve strong performance on many reasoning tasks when allowed to externalize intermediate steps as Chain-of-Thought (CoT). However, many questions require t…
NormAct: Benchmarking Embodied Agents' Proactive Compliance with Unspoken Social Norms
Shiyun Zhao, Xinwei Song, Tianyu Guo +7
Embodied agents driven by multimodal large language models (MLLMs) can often complete everyday tasks from visual observations, but goal achievement does not establish whether they…
Scaling Reasoning Tokens via RL and Parallel Thinking: Evidence From Competitive Programming
Qianfan Zhang, Tianyu Guo, Xuandi Ren +4
We study how to scale reasoning token budgets for competitive programming through two complementary approaches: training-time reinforcement learning (RL) and test-time parallel thi…
TongSIM: A General Platform for Simulating Intelligent Machines
Zhe Sun, Kunlun Wu, Chuanjian Fu +24
As artificial intelligence (AI) rapidly advances, especially in multimodal large language models (MLLMs), research focus is shifting from single-modality text processing to the mor…
SPEED-Q: Staged Processing with Enhanced Distillation towards Efficient Low-bit On-device VLM Quantization
Tianyu Guo, Shanwei Zhao, Shiai Zhu +1
Deploying Vision-Language Models (VLMs) on edge devices (e.g., smartphones and robots) is crucial for enabling low-latency and privacy-preserving intelligent applications. Given th…
GSM-Agent: Understanding Agentic Reasoning Using Controllable Environments
Hanlin Zhu, Tianyu Guo, Song Mei +4
As LLMs are increasingly deployed as agents, agentic reasoning - the ability to combine tool use, especially search, and reasoning - becomes a critical skill. However, it is hard t…
IDA-Bench: Evaluating LLMs on Interactive Guided Data Analysis
Hanyu Li, Haoyu Liu, Tingyu Zhu +4
Large Language Models (LLMs) show promise as data analysis agents, but existing benchmarks overlook the iterative nature of the field, where experts' decisions evolve with deeper i…
Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers
Yixiao Huang, Hanlin Zhu, Tianyu Guo +5
Large language models (LLMs) can acquire new knowledge through fine-tuning, but this process exhibits a puzzling duality: models can generalize remarkably from new facts, yet are a…
How Do LLMs Perform Two-Hop Reasoning in Context?
Tianyu Guo, Hanlin Zhu, Ruiqi Zhang +4
``Socrates is human. All humans are mortal. Therefore, Socrates is mortal.'' This form of argument illustrates a typical pattern of two-hop reasoning. Formally, two-hop reasoning r…
Retrieval-Augmented Generation by Evidence Retroactivity in LLMs
Liang Xiao, Wen Dai, Shuai Chen +4
Retrieval-augmented generation has gained significant attention due to its ability to integrate relevant external knowledge, enhancing the accuracy and reliability of the LLMs' res…
Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs
Tianyu Guo, Druv Pai, Yu Bai +3
Practitioners have consistently observed three puzzling phenomena in transformer-based large language models (LLMs): attention sinks, value-state drains, and residual-state peaks,…
Collaborative Heterogeneous Causal Inference Beyond Meta-analysis
Tianyu Guo, Sai Praneeth Karimireddy, Michael I. Jordan
Collaboration between different data centers is often challenged by heterogeneity across sites. To account for the heterogeneity, the state-of-the-art method is to re-weight the co…
How Do Transformers Learn In-Context Beyond Simple Functions? A Case Study on Learning with Representations
Tianyu Guo, Wei Hu, Song Mei +4
While large language models based on the transformer architecture have demonstrated remarkable in-context learning (ICL) capabilities, understandings of such capabilities are still…