7 citations · 39 across the 30 of their papers we have counts for
41 papers
Compositional RL: Evaluating and Training LLMs for Collaborative Reasoning
Shangbin Feng, Yulia Tsvetkov, Jan Kautz +1
Language models are often evaluated alone but increasingly deployed together, routed between, made to debate, ensembled, or collaboratively trained. For a model inside a system, wh…
Reading, Not Manipulating: Leveraging Router Logits for Multimodal Safety in MoE Vision-Language Models
Ziyuan Yang, Wenxuan Ding, Shangbin Feng +1
Vision-language models (VLMs) face compositional safety risks where harmful intent emerges from the interaction between visual and textual inputs. As mixture-of-experts (MoE) VLMs…
Multi-LLM Collaborative Alignment via Stackelberg Games
Christina Hahn, Shangbin Feng, Dean Light +3
A pool of language models can collaborate and improve collectively by learning from one another's responses. These interactions depend on the instructions used during training. Exi…
You're Hired: Strategic Model Selection for LLM Collaboration
Zongwan Cao, Ziyuan Yang, Shangbin Feng +5
While multi-agent and model collaboration algorithms gain traction to combine the strengths of diverse Large Language Models (LLMs), existing systems remain bottlenecked on pre-def…
Save Your Saturated Data: Learning Beyond Reward Saturation in Group-Based RL
Ziyuan Yang, Yike Wang, Shangbin Feng +1
Group-relative reinforcement learning (RL) relies on reward variation among sampled responses to estimate informative relative advantages. As language models become increasingly ca…
Before the Rollout Ends: Early Terminal Reward Prediction for Long-horizon Coding Agents
Jihan Yao, Sihan Zeng, Shangbin Feng +3
Long-horizon coding agents receive verifiable rewards only after completing expensive sequences of tool calls. This increases inference cost, amplifies early wrong hypotheses, and…
Rethinking the Evaluation of Harness Evolution for Agents
Yike Wang, Huaisheng Zhu, Zhengyu Hu +8
Harness evolution is an iterative search procedure that repeatedly evaluates and revises candidate harnesses used for LLM agents using task feedback. We revisit the evaluation of s…
The Single-Multi Evolution Loop for Self-Improving Model Collaboration Systems
Shangbin Feng, Kishan Panaganti, Yulia Tsvetkov +1
Model collaboration -- systems where multiple language models (LMs) collaborate -- combines the strengths of diverse models with cost in loading multiple LMs. We improve efficiency…
Among Us: Measuring and Mitigating Malicious Contributions in Model Collaboration Systems
Ziyuan Yang, Wenxuan Ding, Shangbin Feng +1
Language models (LMs) are increasingly used in collaboration: multiple LMs trained by different parties collaborate through routing systems, multi-agent debate, model merging, and…
MentorCollab: Large-to-Small Inference-Time Mentorship for Concise Reasoning in Language Models
Haojin Wang, Yike Wang, Shangbin Feng +2
Large reasoning models (LRMs) have demonstrated impressive reasoning capabilities, but their solutions are often verbose and computationally expensive, and taxing for users to read…
Don't Throw Away Your Pretrained Model
Shangbin Feng, Wenhao Yu, Yike Wang +3
Alignment training has tradeoffs: it helps language models (LMs) gain in reasoning and instruction following but might lose out on skills such as creativity and calibration, where…
Interactive Reasoning: Visualizing and Controlling Chain-of-Thought Reasoning in Large Language Models
Rock Yuren Pang, K. J. Kevin Feng, Shangbin Feng +5
The output quality of large language models (LLMs) can be improved via "reasoning": generating segments of chain-of-thought (CoT) content to further condition the model prior to pr…
Escaping the SpuriVerse: Can Large Vision-Language Models Generalize Beyond Seen Spurious Correlations?
Yiwei Yang, Chung Peng Lee, Shangbin Feng +5
Finetuning can cause spurious correlations to arise between non-essential features and the target labels, but benchmarks to study these effects involve contrived settings and narro…
GuessBench: Sensemaking Multimodal Creativity in the Wild
Zifeng Zhu, Shangbin Feng, Herun Wan +3
We propose GuessBench, a novel benchmark that evaluates Vision Language Models (VLMs) on modeling the pervasive, noisy, and pluralistic human creativity. GuessBench sources data fr…
Data Swarms: Optimizable Generation of Synthetic Evaluation Data
Shangbin Feng, Yike Wang, Weijia Shi +1
We propose Data Swarms, an algorithm to optimize the generation of synthetic evaluation data and advance quantitative desiderata of LLM evaluation. We first train a swarm of initia…
SPARTA ALIGNMENT: Collectively Aligning Multiple Language Models through Combat
Yuru Jiang, Wenxuan Ding, Shangbin Feng +2
We propose SPARTA ALIGNMENT, an algorithm to collectively align multiple LLMs through competition and combat. To complement a single model's lack of diversity in generation and bia…
Generalizable LLM Learning of Graph Synthetic Data with Post-training Alignment
Yizhuo Zhang, Heng Wang, Shangbin Feng +3
Previous research has sought to enhance the graph reasoning capabilities of LLMs by supervised fine-tuning on synthetic graph data. While these led to specialized LLMs better at so…
MMMG: a Comprehensive and Reliable Benchmark for Multitask Multimodal Generation
Jihan Yao, Yushi Hu, Yujie Yi +10
Automatically evaluating multimodal generation presents a significant challenge, as automated metrics often struggle to align with human evaluation, especially for complex tasks th…
ScienceMeter: Tracking Scientific Knowledge Updates in Language Models
Yike Wang, Shangbin Feng, Yulia Tsvetkov +1
Large Language Models (LLMs) are increasingly used to support scientific research, but their knowledge of scientific advancements can quickly become outdated. We introduce ScienceM…
FACTS&EVIDENCE: An Interactive Tool for Transparent Fine-Grained Factual Verification of Machine-Generated Text
Varich Boonsanong, Vidhisha Balachandran, Xiaochuang Han +3
With the widespread consumption of AI-generated content, there has been an increased focus on developing automated tools to verify the factual accuracy of such content. However, pr…
When One LLM Drools, Multi-LLM Collaboration Rules
Shangbin Feng, Wenxuan Ding, Alisa Liu +10
This position paper argues that in many realistic (i.e., complex, contextualized, subjective) scenarios, one LLM is not enough to produce a reliable output. We challenge the status…
Heterogeneous Swarms: Jointly Optimizing Model Roles and Weights for Multi-LLM Systems
Shangbin Feng, Zifeng Wang, Palash Goyal +8
We propose Heterogeneous Swarms, an algorithm to design multi-LLM systems by jointly optimizing model roles and weights. We represent multi-LLM systems as directed acyclic graphs (…
Varying Shades of Wrong: Aligning LLMs with Wrong Answers Only
Jihan Yao, Wenxuan Ding, Shangbin Feng +2
In the absence of abundant reliable annotations for challenging tasks and contexts, how can we expand the frontier of LLM capabilities with potentially wrong answers? We focus on t…
Model Swarms: Collaborative Search to Adapt LLM Experts via Swarm Intelligence
Shangbin Feng, Zifeng Wang, Yike Wang +9
We propose Model Swarms, a collaborative search algorithm to adapt LLMs via swarm intelligence, the collective behavior guiding individual systems. Specifically, Model Swarms start…