forum id stringlengths 10 10 | title stringlengths 21 154 | scores listlengths 3 8 | text stringlengths 48.3k 238k |
|---|---|---|---|
MKEHCx25xp | WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild | [
8,
6,
8
] | Published as a conference paper at ICLR 2025
WILDBENCH: BENCHMARKING LLMS WITH
CHALLENGING TASKS FROM REAL USERS IN THE WILD
Bill Yuchen Lin♡♢
Yuntian Deng♡ Khyathi Chandu♡ Faeze Brahman♡ Abhilasha Ravichander♡
Valentina Pyatkin♡ Nouha Dziri♡ Ronan Le Bras♡ Yejin Choi♡♢
♡Allen Institute for AI
♢University of Washin... |
lgsyLSsDRe | NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models | [
8,
8,
6,
8
] | Published as a conference paper at ICLR 2025
NV-EMBED: IMPROVED TECHNIQUES FOR TRAINING
LLMS AS GENERALIST EMBEDDING MODELS
Chankyu Lee ∗ 1
Rajarshi Roy 1
Mengyao Xu 1
Jonathan Raiman 1
Mohammad Shoeybi 1
Bryan Catanzaro 1
Wei Ping ∗ 1
NVIDIA
ABSTRACT
Decoder-only large language model (LLM)-based embedding m... |
et5l9qPUhm | Strong Model Collapse | [
8,
8,
8
] | Published as a conference paper at ICLR 2025
STRONG MODEL COLLAPSE
Elvis Dohmatob1,2,3, Yunzhen Feng4,†, Arjun Subramonian5,†, Julia Kempe1,4
1Meta FAIR 2Concordia University
4NYU 5UCLA
3Mila
†Work done while interning at Meta. Correspondence to elvis.dohmatob@concordia.ca
ABSTRACT
Within the scaling laws paradi... |
8m7p4k6Zeb | From Artificial Needles to Real Haystacks: Improving Retrieval Capabilities in LLMs by Finetuning on Synthetic Data | [
6,
6,
6
] | Published as a conference paper at ICLR 2025
FROM ARTIFICIAL NEEDLES TO REAL HAYSTACKS: IM-
PROVING RETRIEVAL CAPABILITIES IN LLMS BY FINE-
TUNING ON SYNTHETIC DATA
Zheyang Xiongw, Vasilis Papageorgiouw, Kangwook Leew, Dimitris Papailiopoulosw,ms
wUniversity of Wisconsin-Madison, msMicrosoft Research
ABSTRACT
Rece... |
hTphfqtafO | Large Language Models are Interpretable Learners | [
5,
6,
8
] | Published as a conference paper at ICLR 2025
LARGE LANGUAGE MODELS ARE INTERPRETABLE
LEARNERS
Ruochen Wang∗
UCLA
Si Si
Google
Felix Yu
Google
Dorothea Wiesmann
Google
Cho-Jui Hsieh
Google, UCLA
Inderjit Dhillon
Google
ABSTRACT
The trade-off between expressiveness and interpretability remains a core challenge
w... |
kxnoqaisCT | Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents | [
8,
8,
5,
10
] | Published as a conference paper at ICLR 2025
Navigating the Digital World as Humans Do:
UNIVERSAL VISUAL GROUNDING FOR GUI AGENTS
Boyu Gou1 Ruohan Wang1 Boyuan Zheng1 Yanan Xie2 Cheng Chang2 Yiheng Shu1
Huan Sun1 Yu Su1
1The Ohio State University
{gou.43, sun.397, su.809}@osu.edu, yanan@orby.ai
https://osu-nlp-group.... |
590yfqz1LE | Measuring Non-Adversarial Reproduction of Training Data in Large Language Models | [
6,
5,
8,
8,
8,
6,
5,
8
] | Published as a conference paper at ICLR 2025
MEASURING NON-ADVERSARIAL REPRODUCTION
OF TRAINING DATA IN LARGE LANGUAGE MODELS
Michael Aerni∗
Nicholas Carlini2
1
Javier Rando∗
1
Daphne Ippolito2,3
Edoardo Debenedetti1
Florian Tramèr1
1ETH Zurich
2Google DeepMind 3Carnegie Mellon University
ABSTRACT
Large lang... |
FpiCLJrSW8 | More RLHF, More Trust? On The Impact of Preference Alignment On Trustworthiness | [
8,
8,
6,
6
] | Published as a conference paper at ICLR 2025
MORE RLHF, MORE TRUST? ON THE IMPACT OF PREF-
ERENCE ALIGNMENT ON TRUSTWORTHINESS
Aaron J. Li∗, Satyapriya Krishna, Himabindu Lakkaraju
Harvard University
ABSTRACT
The trustworthiness of Large Language Models (LLMs) refers to the extent to
which their outputs are reliabl... |
AqfUa08PCH | Training Language Models on Synthetic Edit Sequences Improves Code Synthesis | [
6,
8,
6,
6
] | Published as a conference paper at ICLR 2025
TRAINING LANGUAGE MODELS ON SYNTHETIC
EDIT SEQUENCES IMPROVES CODE SYNTHESIS
Ulyana Piterbarg, Lerrel Pinto, & Rob Fergus∗
New York University
ABSTRACT
Software engineers mainly write code by editing existing programs. In contrast,
language models (LMs) autoregressively ... |
vhPE3PtTgC | SWEb: A Large Web Dataset for the Scandinavian Languages | [
8,
6,
6,
5
] | Published as a conference paper at ICLR 2025
SWEB: A LARGE WEB DATASET FOR THE
SCANDINAVIAN LANGUAGES
Tobias Norlund∗, Tim Isbister, Amaru Cuba Gyllensten, Paul Dos Santos,
Danila Petrelli, Ariel Ekgren, Magnus Sahlgren
AI Sweden
ABSTRACT
This paper presents the hitherto largest pretraining dataset for the Scandina... |
3c4zQpIFNK | LIME: LESS IS MORE FOR MLLM EVALUATION | [
5,
5,
8,
6
] | Under review as a conference paper at ICLR 2025
LIME: LESS IS MORE FOR MLLM EVALUATION
Anonymous authors
Paper under double-blind review
ABSTRACT
Multimodal Large Language Models (MLLMs) are measured on numerous bench-
marks like image captioning, visual question answer, and reasoning. However,
these benchmarks oft... |
KmmNb7631I | Learning to Plan Before Answering: Self-Teaching LLMs to Learn Abstract Plans for Problem Solving | [
6,
5,
8,
6
] | Published as a conference paper at ICLR 2025
LEARNING TO PLAN BEFORE ANSWERING: SELF-
TEACHING LLMS TO LEARN ABSTRACT PLANS FOR
PROBLEM SOLVING
Jin Zhang1,2, Flood Sung2, Zhilin Yang2, Yang Gao1, Chongjie Zhang3
1Institute for Interdisciplinary Information Sciences, Tsinghua University, China
2Moonshot AI
3Washingto... |
0Fi3u4RCyU | Evolve: Evaluating and Optimizing LLMs For Exploration | [
5,
8,
5,
8
] | Under review as a conference paper at ICLR 2025
EVOLVE: EVALUATING AND OPTIMIZING LLMS FOR
EXPLORATION
Anonymous authors
Paper under double-blind review
ABSTRACT
Despite their success in many domains, large language models (LLMs) remain
under-studied in scenarios requiring optimal decision-making under uncertainty.... |
IDJUscOjM3 | Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
SELF-MOE: TOWARDS COMPOSITIONAL LARGE LAN-
GUAGE MODELS WITH SELF-SPECIALIZED EXPERTS
Junmo Kang∗
Georgia Tech
Leonid Karlinsky
MIT-IBM Watson AI Lab
Hongyin Luo
MIT
Zhen Wang
UCSD
Jacob Hansen
MIT
James Glass
MIT
David Cox
MIT-IBM Watson AI Lab
Rameswar Panda
MIT-I... |
o9ewXD1JuB | OLAPH: Improving Factuality in Biomedical Long-form Question Answering | [
5,
6,
8,
6
] | Under review as a conference paper at ICLR 2025
OLAPH:
LONG-FORM QUESTION ANSWERING
IMPROVING FACTUALITY IN BIOMEDICAL
Anonymous authors
Paper under double-blind review
ABSTRACT
In the medical domain, numerous scenarios necessitate the long-form generation
ability of large language models (LLMs). Specifically, whe... |
syThiTmWWm | Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates | [
8,
10,
10,
8,
6,
6,
6,
8
] | Published as a conference paper at ICLR 2025
CHEATING AUTOMATIC LLM BENCHMARKS:
NULL MODELS ACHIEVE HIGH WIN RATES
Xiaosen Zheng∗1,2, Tianyu Pang∗†1, Chao Du1, Qian Liu1, Jing Jiang†2,3, Min Lin1
1Sea AI Lab, Singapore
2Singapore Management University
{zhengxs, tianyupang, duchao, liuqian, linmin}@sea.com;
jingjiang@... |
wFs2E5wCw6 | Tree of Attributes Prompt Learning for Vision-Language Models | [
6,
6,
6,
6,
8
] | Published as a conference paper at ICLR 2025
TREE OF ATTRIBUTES PROMPT LEARNING FOR VISION-
LANGUAGE MODELS
Tong Ding1,2 Wanhua Li1∗ Zhongqi Miao3 Hanspeter Pfister1
1Harvard University 2Mass General Brigham 3Microsoft
ABSTRACT
Prompt learning has proven effective in adapting vision language models for
downstream t... |
sfQ6XpApfS | PiCO: Peer Review in LLMs based on Consistency Optimization | [
6,
6,
6
] | Published as a conference paper at ICLR 2025
PICO: P EER REVIEW IN LLM S BASED ON CONSIS -
TENCY OPTIMIZATION
Kun-Peng Ning1, Shuo Yang1, Yu-Yang Liu1,∗, Jia-Yu Yao1, Zhen-Hui Liu 1,
Yong-Hong Tian1,2, Yibing Song, Li Yuan1,2,∗
1School of Electrical and Computer Engineering, Peking University
2Peng Cheng Laboratory
{ni... |
VOAMTA8jKu | DynaMath: A Dynamic Visual Benchmark for Evaluating Mathematical Reasoning Robustness of Vision Language Models | [
6,
6,
8,
8
] | Published as a conference paper at ICLR 2025
DYNAMATH: A DYNAMIC VISUAL BENCHMARK
FOR EVALUATING MATHEMATICAL REASONING
ROBUSTNESS OF VISION LANGUAGE MODELS
Chengke Zou1,2∗ †, Xingang Guo1∗, Rui Yang1∗, Junyu Zhang1, Bin Hu1, Huan Zhang1
1University of Illinois at Urbana-Champaign, 2University of California, Berkele... |
vPOMTkmSiu | Scaling Laws for Downstream Task Performance in Machine Translation | [
3,
6,
8,
8,
8
] | Published as a conference paper at ICLR 2025
SCALING LAWS FOR DOWNSTREAM TASK
PERFORMANCE IN MACHINE TRANSLATION
Berivan Isik♣, Natalia Ponomareva♣, Hussein Hazimeh♦∗, Dimitris Paparas♣
Sergei Vassilvitskii♣, Sanmi Koyejo§∗
♣Google Research, ♦OpenAI, §Stanford University
berivan@google.com
ABSTRACT
Scaling laws pr... |
BkwCrIsTbR | Scaling Instruction-tuned LLMs to Million-token Contexts via Hierarchical Synthetic Data Generation | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
SCALING INSTRUCTION-TUNED LLMS TO MILLION-
TOKEN CONTEXTS VIA HIERARCHICAL SYNTHETIC
DATA GENERATION
Linda He1∗ Jue Wang2 Maurice Weber2 Shang Zhu2 Ben Athiwaratkun2 Ce Zhang2,3
1Harvard University 2Together AI
3University of Chicago
lindahe@college.harvard.edu, {jue, maur... |
8EB8k6DdCU | ToolACE: Enhancing Function Calling with Accuracy, Complexity, and Diversity | [
6,
6,
6,
8
] | Published as a conference paper at ICLR 2025
TOOLACE: WINNING THE POINTS OF LLM FUNCTION CALLING
Weiwen Liu†1, Xu Huang†3, Xingshan Zeng†2, Xinlong Hao2, Shuai Yu2, Dexun Li2, Shuai Wang2,
Weinan Gan2, Zhengying Liu2, Yuanqing Yu5, Zezhong Wang6, Yuxian Wang4, Wu Ning4, Yutai
Hou4, Bin Wang2, Chuhan Wu*2, Xinzhi Wang... |
jki6EFsZLw | OmnixR: Evaluating Omni-modality Language Models on Reasoning across Modalities | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
OMNI R: EVALUATING OMNI-MODALITY LANGUAGE
MODELS ON REASONING ACROSS MODALITIES
Lichang Chen12 ∗, Hexiang Hu1, Mingda Zhang1, Yiwen Chen1, Zifeng Wang1, Yandong Li1,
Pranav Shyam1, Tianyi Zhou2, Heng Huang2, Ming-Hsuan Yang1, Boqing Gong1
Google DeepMind1; University of Ma... |
UsRKFYR4lM | Mitigating Spurious Correlations in Zero-Shot Multimodal Models | [
6,
6,
8,
6
] | Published as a conference paper at ICLR 2025
MITIGATING SPURIOUS CORRELATIONS IN ZERO-
SHOT MULTIMODAL MODELS
Shenyu Lu, Junyi Chai & Xiaoqian Wang∗
Elmore Family School of Electrical and Computer Engineering
Purdue University
West Lafayette, IN 47906, USA
{lu876,chai28,joywang}@purdue.edu
ABSTRACT
Multimodal model... |
jjCB27TMK3 | Data Mixing Laws: Optimizing Data Mixtures by Predicting Language Modeling Performance | [
8,
6,
6,
6
] | Published as a conference paper at ICLR 2025
DATA MIXING LAWS: OPTIMIZING DATA MIXTURES BY
PREDICTING LANGUAGE MODELING PERFORMANCE
Jiasheng Ye1∗, Peiju Liu1∗, Tianxiang Sun1, Jun Zhan1, Yunhua Zhou2†, Xipeng Qiu1†
1Fudan University, 2Shanghai AI Labortory
{jsye23,pjliu23}@m.fudan.edu.cn
zhouyunhua@pjlab.org.cn xpqiu... |
EDoD3DgivF | On Linear Representations and Pretraining Data Frequency in Language Models | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
ON LINEAR REPRESENTATIONS AND PRETRAINING
DATA FREQUENCY IN LANGUAGE MODELS
Jack Merullo♢ Noah A. Smith♡♣
Sarah Wiegreffe∗♡♣ Yanai Elazar∗♡♣
♢Brown University, ♡Allen Institute for AI (Ai2), ♣University of Washington
∗Co-senior authors.
jack merullo@brown.edu, {noah, sar... |
zP8HygcAMY | Can LLMs Evaluate Complex Attribution in QA? Automatic Benchmarking Using Knowledge Graphs | [
6,
6,
5,
6
] | Under review as a conference paper at ICLR 2025
CAN LLMS EVALUATE COMPLEX ATTRIBUTION IN
QA? AUTOMATIC BENCHMARKING USING KNOWL-
EDGE GRAPHS
Anonymous authors
Paper under double-blind review
ABSTRACT
The attribution of question answering (QA), which is to get evidences for sup-
porting the generated answer, has att... |
ybfmpJiKXX | AIMS.au: A Dataset for the Analysis of Modern Slavery Countermeasures in Corporate Statements | [
6,
8,
5
] | Published as a conference paper at ICLR 2025
AIMS.AU: A DATASET FOR THE ANALYSIS OF
MODERN SLAVERY COUNTERMEASURES IN
CORPORATE STATEMENTS
Adriana Eufrosiana Bora1,2
Pierre-Luc St-Charles1 Mirko Bronzi1
Arsène Fansi Tchango1 Bruno Rousseau1 Kerrie Mengersen2
1Mila - Quebec AI Institute
{adriana.eufrosina-bora, pl.s... |
iv1TpRCJeK | $\forall$uto$\exists$$\lor\!\land$L: Autonomous Evaluation of LLMs for Truth Maintenance and Reasoning Tasks | [
8,
6,
5
] | Published as a conference paper at ICLR 2025
AUTOEVAL: AUTONOMOUS EVALUATION OF LLMS
FOR TRUTH MAINTENANCE AND REASONING TASKS
Rushang Karia∗, Daniel Bramblett∗, Daksh Dobhal, Siddharth Srivastava
School of Computing and Augmented Intelligence
Arizona State University
{rushang.karia,drbrambl,ddobhal,siddharths}@asu.e... |
suz4utPr9Y | How efficient is LLM-generated code? A rigorous & high-standard benchmark | [
6,
5,
6,
6
] | Published as a conference paper at ICLR 2025
HOW EFFICIENT IS LLM-GENERATED CODE?
A RIGOROUS & HIGH-STANDARD BENCHMARK
Ruizhong Qiu†, Weiliang Will Zeng‡, James Ezick‡, Christopher Lott‡, & Hanghang Tong†
†University of Illinois Urbana–Champaign
{rq5,htong}@illinois.edu
{wzeng,jezick,clott}@qti.qualcomm.com
‡Qualco... |
huuKoVQnB0 | Improving Pretraining Data Using Perplexity Correlations | [
6,
5,
8,
5,
6
] | Published as a conference paper at ICLR 2025
IMPROVING PRETRAINING DATA USING
PERPLEXITY CORRELATIONS
Tristan Thrush, Christopher Potts & Tatsunori Hashimoto
Department of Computer Science
Stanford University
Stanford, CA 94305, USA
{tthrush,cgpotts,thashim}@stanford.edu
ABSTRACT
Quality pretraining data is oft... |
jlzNb1iWs3 | The OMG dataset: An Open MetaGenomic corpus for mixed-modality genomic language modeling | [
5,
6,
8,
5
] | Published as a conference paper at ICLR 2025
The OMG dataset: An Open MetaGenomic
corpus for mixed-modality genomic language
modeling
Andre Cornman*,1
Simon Roux2 Martin Beracochea3 Milot Mirdita4
Jacob West-Roberts1
Antonio Pedro Camargo2
Sergey Ovchinnikov5
Yunha Hwang*,1
1Tatta Bio, USA
2DOE Joint Genome Insti... |
3UKOzGWCVY | Learn-by-interact: A Data-Centric Framework For Self-Adaptive Agents in Realistic Environments | [
6,
8,
6,
6
] | Published as a conference paper at ICLR 2025
LEARN-BY-INTERACT: A DATA-CENTRIC FRAME-
WORK FOR SELF-ADAPTIVE AGENTS IN REALISTIC
ENVIRONMENTS
Hongjin Su 12 , Ruoxi Sun 1 , Jinsung Yoon 1 , Pengcheng Yin 1 , Tao Yu 2 , Sercan Ö. Arık 1
1 Google , 2 The University of Hong Kong
ABSTRACT
Autonomous agents powered by la... |
womU9cEwcO | Autonomous agents from automatic reward modeling and planning | [
6,
6,
8
] | Published as a conference paper at ICLR 2025
ARMAP: SCALING AUTONOMOUS AGENTS VIA
AUTOMATIC REWARD MODELING AND PLANNING
Zhenfang Chen∗
MIT-IBM Watson AI Lab
Delin Chen∗
UMass Amherst
Rui Sun∗
University of California, Los Angeles
Wenjun Liu∗
UMass Amherst
Chuang Gan
UMass Amherst and MIT-IBM Watson AI Lab
ABSTR... |
V892sBHUbN | Rapid Response: Mitigating LLM Jailbreaks With A Few Examples | [
5,
8,
5,
5
] | Under review as a conference paper at ICLR 2025
RAPID RESPONSE: MITIGATING LLM JAILBREAKS
WITH A FEW EXAMPLES
Anonymous authors
Paper under double-blind review
ABSTRACT
As large language models (LLMs) grow more powerful, ensuring their safety
against misuse becomes crucial. While researchers have focused on develop... |
QxbJYBZVbE | CursorCore: Assist Programming through Aligning Anything | [
8,
5,
6,
5
] | Under review as a conference paper at ICLR 2025
CURSORCORE: ASSIST PROGRAMMING
THROUGH ALIGNING ANYTHING
Anonymous authors
Paper under double-blind review
ABSTRACT
Large language models have been successfully applied to programming assistance
tasks, such as code completion, code insertion, and instructional code ed... |
OQqNieeivq | KaSA: Knowledge-Aware Singular-Value Adaptation of Large Language Models | [
6,
8,
8,
5,
6
] | Published as a conference paper at ICLR 2025
KASA: KNOWLEDGE-AWARE
ADAPTATION OF LARGE LANGUAGE MODELS
SINGULAR-VALUE
Fan Wang∗♡, Juyong Jiang∗♡, Chansung Park∗♠, Sunghun Kim†♡♣, Jing Tang†♡♣
♡The Hong Kong University of Science and Technology (Guangzhou)
♠Electronics and Telecommunications Research Institute
♣The H... |
IHRQif8VQC | Ensemble everything everywhere: Multi-scale aggregation for adversarial robustness | [
5,
8,
8,
6
] | Under review as a conference paper at ICLR 2025
ENSEMBLE EVERYTHING EVERYWHERE: MULTI-
SCALE AGGREGATION FOR ADVERSARIAL ROBUST-
NESS
Anonymous authors
Paper under double-blind review
ABSTRACT
Adversarial examples pose a significant challenge to the robustness, reliability and
alignment of deep neural networks. We p... |
3OyaXFQuDl | Smaller, Weaker, Yet Better: Training LLM Reasoners via Compute-Optimal Sampling | [
8,
8,
6,
6
] | Published as a conference paper at ICLR 2025
SMALLER, WEAKER, YET BETTER: TRAINING LLM
REASONERS VIA COMPUTE-OPTIMAL SAMPLING
Hritik Bansal1,2, Arian Hosseini1,3, Rishabh Agarwal1,3, Vinh Q. Tran1, Mehran Kazemi1 ∗
1 Google DeepMind, 2 UCLA, 3 Mila
Correspondence: hbansal@g.ucla.edu and mehrankazemi@google.com
ABSTR... |
cFu7ze7xUm | DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads | [
8,
5,
5,
6,
8,
6
] | Published as a conference paper at ICLR 2025
DUOATTENTION: EFFICIENT LONG-CONTEXT LLM
INFERENCE WITH RETRIEVAL AND STREAMING HEADS
Guangxuan Xiao1 ∗ Jiaming Tang1
Shang Yang1 Haotian Tang1 Yao Fu4
1 MIT
https://github.com/mit-han-lab/duo-attention
2 Tsinghua University
3 SJTU 4University of Edinburgh
Jingwei Zuo2
... |
qssVptHTPN | Locality Alignment Improves Vision-Language Models | [
5,
6,
5,
8
] | Published as a conference paper at ICLR 2025
LOCALITY ALIGNMENT IMPROVES VISION-LANGUAGE
MODELS
Ian Covert, Tony Sun, James Zou∗, Tatsunori Hashimoto∗
Stanford University
{icovert, suntony, jamesz, thashim}@stanford.edu
ABSTRACT
Vision language models (VLMs) have seen growing adoption in recent years,
but many stil... |
s5epFPdIW6 | MMed-RAG: Versatile Multimodal RAG System for Medical Vision Language Models | [
6,
6,
8,
8
] | Published as a conference paper at ICLR 2025
MMED-RAG: VERSATILE MULTIMODAL RAG SYS-
TEM FOR MEDICAL VISION LANGUAGE MODELS
Peng Xia1, Kangyu Zhu2, Haoran Li3, Tianze Wang4, Weijia Shi5, Sheng Wang5,
Linjun Zhang4, James Zou6, Huaxiu Yao1
1UNC-Chapel Hill, 2Brown University, 3Carnegie Mellon University, 4Rutgers Univ... |
ymt4crbbXh | AutoBencher: Towards Declarative Benchmark Construction | [
5,
6,
8,
6
] | Published as a conference paper at ICLR 2025
AUTOBENCHER: TOWARDS DECLARATIVE BENCHMARK
CONSTRUCTION
Xiang Lisa Li, Farzaan Kaiyom, Evan Zheran Liu, Yifan Mai, Percy Liang, Tatsunori Hashimoto
Stanford University
xlisali@stanford.edu
ABSTRACT
We present AutoBencher, a declarative framework for automatic benchmark c... |
UxkznlcnHf | Towards a Theoretical Understanding of Synthetic Data in LLM Post-Training: A Reverse-Bottleneck Perspective | [
3,
8,
8,
6
] | Published as a conference paper at ICLR 2025
TOWARDS A THEORETICAL UNDERSTANDING OF SYN-
THETIC DATA IN LLM POST-TRAINING:
A REVERSE-BOTTLENECK PERSPECTIVE
Zeyu Gan, Yong Liu∗
Gaoling School of Artificial Intelligence
Renmin University of China
Beijing, China
{zygan,liuyonggsai}@ruc.edu.cn
ABSTRACT
Synthetic data h... |
qn9tBYQHGi | Do LLM Agents Have Regret? A Case Study in Online Learning and Games | [
6,
8,
6,
6
] | Published as a conference paper at ICLR 2025
DO LLM AGENTS HAVE REGRET?
A CASE STUDY IN ONLINE LEARNING AND GAMES
Chanwoo Parkω 1, Xiangyu Liuω2, Asuman Ozdaglar1, Kaiqing Zhang2
1 MIT, 2 University of Maryland, College Park
ABSTRACT
Large language models (LLMs) have been increasingly employed for (interac-
tive) d... |
mPdmDYIQ7f | AgentSquare: Automatic LLM Agent Search in Modular Design Space | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
AGENTSQUARE: AUTOMATIC LLM AGENT
SEARCH IN MODULAR DESIGN SPACE
Yu Shang1∗, Yu Li2∗, Keyu Zhao1, Likai Ma1, Jiahe Liu1, Fengli Xu1†, Yong Li1†
1Department of Electronic Engineering, Tsinghua University
2Shenzhen International Graduate School, Tsinghua University
{fenglixu... |
QoDDNkx4fP | ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference-Time | [
8,
5,
5,
6
] | Published as a conference paper at ICLR 2025
ETA: EVALUATING THEN ALIGNING SAFETY OF VI-
SION LANGUAGE MODELS AT INFERENCE TIME
Yi Ding, Bolian Li, Ruqi Zhang
Department of Computer Science, Purdue University, USA
{ding432,li4468,ruqiz}@purdue.edu
ABSTRACT
Vision Language Models (VLMs) have become essential backbon... |
7NL74jUiMg | Alchemy: Amplifying Theorem-Proving Capability Through Symbolic Mutation | [
6,
6,
6,
8
] | Published as a conference paper at ICLR 2025
ALCHEMY : A MPLIFYING THEOREM -PROVING CAPA-
BILITY THROUGH SYMBOLIC MUTATION
Shaonan Wu1,2, ∗ Shuai Lu3,† Yeyun Gong3, Nan Duan3, Ping Wei1,2,†
1 National Key Laboratory of Human-Machine Hybrid Augmented Intelligence
2 Institute of Artificial Intelligence and Robotics, Xi’a... |
I4YU0oECtK | Bayesian scaling laws for in-context learning | [
8,
5,
6,
5
] | 000
001
002
003
004
005
006
007
008
009
010
011
012
013
014
015
016
017
018
019
020
021
022
023
024
025
026
027
028
029
030
031
032
033
034
035
036
037
038
039
040
041
042
043
044
045
046
047
048
049
050
051
052
053
BAYESIAN SCALING LAWS FOR IN-CONTEXT LEARNING
Anonymous authors
Paper under double-blind revi... |
X8dzvdkQwO | Fine-tuning can Help Detect Pretraining Data from Large Language Models | [
5,
6,
8,
6
] | Published as a conference paper at ICLR 2025
FINE-TUNING CAN HELP DETECT PRETRAINING DATA
FROM LARGE LANGUAGE MODELS
Hengxiang Zhang1, Songxin Zhang1, Bingyi Jing1, Hongxin Wei1∗
1Department of Statistics and Data Science, Southern University of Science and Technology
ABSTRACT
In the era of large language models (L... |
Zk9guOl9NS | What Makes Large Language Models Reason in (Multi-Turn) Code Generation? | [
8,
5,
8
] | Published as a conference paper at ICLR 2025
WHAT MAKES LARGE LANGUAGE MODELS REASON
IN (MULTI-TURN) CODE GENERATION?
Kunhao Zheng1,2∗, Juliette Decugis1∗, Jonas Gehring1, Taco Cohen1,
Benjamin Negrevergne2, Gabriel Synnaeve1
1Meta AI (FAIR), 2Paris Dauphine University - PSL
{kunhao, jdecugis, gab}@meta.com
ABSTRACT... |
S85PP4xjFD | ContraFusion: Contrastively Improving Compositional Understanding in Diffusion Models via Fine-Grained Negative Images | [
8,
8,
8,
6
] | Published as a conference paper at ICLR 2025
PROGRESSIVE COMPOSITIONALITY IN
TEXT-TO-IMAGE GENERATIVE MODELS
Evans Xu Han1 Linghao Jin2 Xiaofeng Liu1
1Yale University, 2University of Southern California, 3Massachusetts Institute of Technology
{xu.han.xh365, xiaofeng.liu}@yale.edu
linghaoj@usc.edu
ppliang@mit.edu
Pa... |
tRNKe2Vgqt | MMWorld: Towards Multi-discipline Multi-faceted World Model Evaluation in Videos | [
6,
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
MMWORLD: TOWARDS MULTI-DISCIPLINE MULTI-
FACETED VIDEO UNDERSTANDING EVALUATION
Xuehai He1, Weixi Feng2, Kaizhi Zheng1, Yujie Lu2, Wanrong Zhu2, Jiachen Li2, Yue Fan2,
Jianfeng Wang3, Linjie Li3, Zhengyuan Yang3, Kevin Lin3,
William Yang Wang2, Lijuan Wang3, Xin Eric Wang1... |
1hQKHHUsMx | What Kind of Pretraining Data Do Large Language Models Rely on When Doing Reasoning? | [
6,
8,
8,
5
] | Published as a conference paper at ICLR 2025
PROCEDURAL KNOWLEDGE IN PRETRAINING DRIVES
REASONING IN LARGE LANGUAGE MODELS
Laura Ruis∗
AI Centre, UCL
Maximilian Mozes
Cohere
Juhan Bae
University of Toronto & Vector Institute
Siddhartha Rao Kamalakara
Cohere
Dwarak Talupuru
Cohere
Acyr Locatelli
Cohere
Robert Ki... |
1KvYxcAihR | TMGBench: A Systematic Game Benchmark for Evaluating Strategic Reasoning Abilities of LLMs | [
5,
5,
8,
5
] | Under review as a conference paper at ICLR 2025
TMGBENCH: A SYSTEMATIC GAME BENCHMARK
FOR EVALUATING STRATEGIC REASONING ABILITIES
OF LLMS
Anonymous authors
Paper under double-blind review
ABSTRACT
The rapid advancement of large language models (LLMs) has accelerated their
application in reasoning, with strategic r... |
cRR0oDFEBC | Self-play with Execution Feedback: Improving Instruction-following Capabilities of Large Language Models | [
6,
8,
8,
8,
6
] | Published as a conference paper at ICLR 2025
SELF-PLAY WITH EXECUTION FEEDBACK: IMPROVING
INSTRUCTION-FOLLOWING CAPABILITIES OF LARGE
LANGUAGE MODELS
Guanting Dong∗, Keming Lu, Chengpeng Li∗, Tingyu Xia∗, Bowen Yu†
Chang Zhou, Jingren Zhou
Qwen Team, Alibaba Inc.
{dongguanting.dgt,lukeming.lkm,lichengpeng.lcp}@alibab... |
636M0nNbPs | Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning | [
6,
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
PAINTING WITH WORDS: ELEVATING DETAILED IM-
AGE CAPTIONING WITH BENCHMARK AND ALIGNMENT
LEARNING
Qinghao Ye*, Xianhan Zeng*, Fu Li, Chunyuan Li, Haoqi Fan
ByteDance Research
ABSTRACT
Image captioning has long been a pivotal task in visual understanding, with recent
advan... |
y3zswp3gek | HarmAug: Effective Data Augmentation for Knowledge Distillation of Safety Guard Models | [
6,
6,
10,
6
] | Published as a conference paper at ICLR 2025
HARMAUG: EFFECTIVE DATA AUGMENTATION FOR
KNOWLEDGE DISTILLATION OF SAFETY GUARD
MODELS
2Theori
3Universit´e de Montr´eal
Seanie Lee1∗ Haebin Seong2∗ Dong Bok Lee1 Minki Kang1 Xiaoyin Chen3,4
Dominik Wagner5 Yoshua Bengio3,4,6
1KAIST
5Technische Hochschule N¨urnberg Georg... |
6RiBl5sCDF | GeoX: Geometric Problem Solving Through Unified Formalized Vision-Language Pre-training | [
6,
8,
6,
8
] | Published as a conference paper at ICLR 2025
GEOX: GEOMETRIC PROBLEM SOLVING THROUGH
UNIFIED FORMALIZED VISION-LANGUAGE PRE-
TRAINING
Renqiu Xia1,2,∗, Mingsheng Li2,3,∗, Hancheng Ye2, Wenjie Wu1, Hongbin Zhou2,
Jiakang Yuan2,3, Tianshuo Peng2,4, Xinyu Cai2, Xiangchao Yan2, Bin Wang2, Conghui He2,
Botian Shi2, Tao Che... |
rawj2PdHBq | Can Medical Vision-Language Pre-training Succeed with Purely Synthetic Data? | [
8,
5,
5
] | Under review as a conference paper at ICLR 2025
CAN MEDICAL VISION-LANGUAGE PRE-TRAINING
SUCCEED WITH PURELY SYNTHETIC DATA?
Anonymous authors
Paper under double-blind review
ABSTRACT
Medical Vision-Language Pre-training (MedVLP) has made significant progress
in enabling zero-shot tasks for medical image understand... |
leSbzBtofH | AutoAdvExBench: Benchmarking Autonomous Exploitation of Adversarial Example Defenses | [
8,
5,
8,
6,
5,
5
] | Under review as a conference paper at ICLR 2025
AUTOADVEXBENCH:
BENCHMARKING AUTONOMOUS EXPLOITATION
OF ADVERSARIAL EXAMPLE DEFENSES
Anonymous authors
Paper under double-blind review
ABSTRACT
We introduce AutoAdvExBench, a benchmark to evaluate if large language mod-
els (LLMs) can autonomously exploit defenses to ... |
44CoQe6VCq | Test of Time: A Benchmark for Evaluating LLMs on Temporal Reasoning | [
8,
6,
8,
6
] | Published as a conference paper at ICLR 2025
TEST OF TIME: A BENCHMARK FOR EVALUATING
LLMS ON TEMPORAL REASONING
Bahare Fatemi1∗, Mehran Kazemi2∗, Anton Tsitsulin1, Karishma Malkan2, Jinyeong Yim3,
John Palowitch2, Sungyong Seo3, Jonathan Halcrow1, and Bryan Perozzi1
1Google Research, 2Google DeepMind, 3Google
ABSTR... |
9QPH1YQCMn | Infilling Score: A Pretraining Data Detection Algorithm for Large Language Models | [
3,
8,
8,
6
] | Published as a conference paper at ICLR 2025
INFILLING SCORE ✼ A PRETRAINING DATA DETECTION
ALGORITHM FOR LARGE LANGUAGE MODELS
Negin Raoof Litu Rout Giannis Daras
Sujay Sanghavi Constantine Caramanis Sanjay Shakkottai Alexandros G. Dimakis
The University of Texas at Austin
{neginmr, litu.rout,giannisdara,constanti... |
nDvgHIBRxQ | Is Your Model Really A Good Math Reasoner? Evaluating Mathematical Reasoning with Checklist | [
8,
6,
6,
5
] | Published as a conference paper at ICLR 2025
IS YOUR MODEL REALLY A GOOD MATH REASONER?
EVALUATING MATHEMATICAL REASONING WITH
CHECKLIST
Zihao Zhou12∗ Shudong Liu3∗ Maizhen Ning126 Wei Liu4
Derek F. Wong3 Xiaowei Huang2 Qiufeng Wang1† Kaizhu Huang6
1Xi’an Jiaotong-liverpool University
4HKUST
https://mathcheck.github.... |
bR1J7SpzrD | Synthio: Augmenting Small-Scale Audio Classification Datasets with Synthetic Data | [
8,
8,
5,
6
] | Published as a conference paper at ICLR 2025
SYNTHIO: AUGMENTING SMALL-SCALE AUDIO CLAS-
SIFICATION DATASETS WITH SYNTHETIC DATA
♦♠∗
Sreyan Ghosh
Dinesh Manocha
♦
♠
, Sonal Kumar
♠
♠∗
, Zhifeng Kong
♦
, Rafael Valle
♦
, Bryan Catanzaro
♦
NVIDIA, CA, USA,
University of Maryland, College Park, USA
{sreyang... |
0bmGL4q7vJ | Multi-modal Agent Tuning: Building a VLM-Driven Agent for Efficient Tool Usage | [
6,
8,
8,
8
] | Published as a conference paper at ICLR 2025
MULTI-MODAL AGENT TUNING: BUILDING A VLM-
DRIVEN AGENT FOR EFFICIENT TOOL USAGE
Zhi Gao1,2∗, Bofei Zhang2∗, Pengxiang Li3,2∗, Xiaojian Ma2, Tao Yuan2, Yue Fan2
Yuwei Wu3,4(cid:0), Yunde Jia4, Song-Chun Zhu1,2,5, Qing Li2(cid:0)
1School of Intelligence Science and Technolo... |
TuOTSAiHDn | MIND: Math Informed syNthetic Dialogues for Pretraining LLMs | [
8,
5,
5
] | MIND
MIND: MATH INFORMED SYNTHETIC DIALOGUES
FOR PRETRAINING LLMS
Syeda Nahida Akter2∗, Shrimai Prabhumoye1,3, John Kamalu1, Sanjeev Satheesh1
Eric Nyberg2, Mostofa Patwary1, Mohammad Shoeybi1, Bryan Catanzaro1
NVIDIA1, Carnegie Mellon University2, Boston University3
sakter@andrew.cmu.edu, sprabhumoye@nvidia.com
ABS... |
IwhvaDrL39 | Research Town: Simulator of Research Community | [
6,
6,
5,
6
] | 000
001
002
003
004
005
006
007
008
009
010
011
012
013
014
015
016
017
018
019
020
021
022
023
024
025
026
027
028
029
030
031
032
033
034
035
036
037
038
039
040
041
042
043
044
045
046
047
048
049
050
051
052
053
RESEARCHTOWN: SIMULATOR OF HUMAN RESEARCH
COMMUNITY
Anonymous authors
Paper under double-blin... |
OZbFRNhpwr | SPA-BENCH: A COMPREHENSIVE BENCHMARK FOR SMARTPHONE AGENT EVALUATION | [
8,
8,
6
] | Published as a conference paper at ICLR 2025
SPA-BENCH: A COMPREHENSIVE BENCHMARK FOR
SMARTPHONE AGENT EVALUATION
Jingxuan Chen1*, Derek Yuen1*, Bin Xie2, Yuhao Yang1, Gongwei Chen2, Zhihao Wu1,
Yixing Li2, Xurui Zhou2, Weiwen Liu1, Shuai Wang1, Kaiwen Zhou1, Rui Shao2†,
Liqiang Nie2, Yasheng Wang1, Jianye Hao1,3, ... |
o5TsWTUSeF | ChartMoE: Mixture of Diversely Aligned Expert Connector for Chart Understanding | [
6,
5,
8,
8
] | Published as a conference paper at ICLR 2025
CHARTMOE: MIXTURE OF DIVERSELY ALIGNED EX-
PERT CONNECTOR FOR CHART UNDERSTANDING
Zhengzhuo Xu12∗ Bowen Qu13∗ Yiyan Qi1∗ Sinan Du2 Chengjin Xu1 Chun Yuan2† Jian Guo14†
1International Digital Economy Academy 2Tsinghua University 3Peking University
4Hong Kong University of ... |
y9A2TpaGsE | Language Agents Meet Causality -- Bridging LLMs and Causal World Models | [
6,
6,
6,
8
] | LANGUAGE AGENTS MEET CAUSALITY – BRIDGING
LLMS AND CAUSAL WORLD MODELS
John Gkountouras1, Matthias Lindemann2, Phillip Lippe3, Efstratios Gavves3,4, and Ivan Titov2,1
1Institute for Logic, Language and Computation (ILLC), University of Amsterdam
2Institute for Language, Cognition and Computation (ILCC), University of... |
KvaDHPhhir | Sketch2Diagram: Generating Vector Diagrams from Hand-Drawn Sketches | [
8,
6,
5,
6
] | Published as a conference paper at ICLR 2025
SKETCH2DIAGRAM: GENERATING VECTOR DIA-
GRAMS FROM HAND-DRAWN SKETCHES
Itsumi Saito*, †, Haruto Yoshida*, Keisuke Sakaguchi*, †
*Tohoku University, †RIKEN AIP
itsumi.saito@tohoku.ac.jp
ABSTRACT
We address the challenge of automatically generating high-quality vector dia-
... |
pXlmOmlHJZ | In-Context Learning of Representations | [
8,
6,
6,
6
] | Published as a conference paper at ICLR 2025
ICLR: IN-CONTEXT LEARNING OF REPRESENTATIONS
Core Francisco Park∗1,2,3 , Andrew Lee∗4, Ekdeep Singh Lubana∗1,3, Yongyi Yang∗1,3,5,
Maya Okawa1,3, Kento Nishi1,4, Martin Wattenberg4, & Hidenori Tanaka1,3
1CBS-NTT Program in Physics of Intelligence, Harvard University
2Depar... |
YrycTjllL0 | BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions | [
8,
8,
10,
10
] | Published as a conference paper at ICLR 2025
BI GCO D EBE N C H: BENCHMARKING CODE GENERA-
TION WITH DIVERSE FUNCTION CALLS AND COMPLEX
INSTRUCTIONS
Junda He3
Imam Nur Bani Yusuf3 Haolan Zhan1
Jenny Chim5 Han Hu1,3 Wenhao Yu12
Terry Yue Zhuo1,2,3 Minh Chien Vu4
Ratnadira Widyasari3
Indraneil Paul7 Simon Brunner8 C... |
v8qABSeeKO | MMKE-Bench: A Multimodal Editing Benchmark for Diverse Visual Knowledge | [
8,
6,
5,
6
] | Published as a conference paper at ICLR 2025
MMKE-BENCH: A MULTIMODAL EDITING BENCH-
MARK FOR DIVERSE VISUAL KNOWLEDGE
Yuntao Du1,2∗, Kailin Jiang3,1∗, Zhi Gao1,4, Chenrui Shi5,1, Zilong Zheng1†, Siyuan Qi1, Qing Li1†
1State Key Laboratory of General Artificial Intelligence, BIGAI
2School of Software & Joint SDU-NTU ... |
F5R0lG74Tu | DataGen: Unified Synthetic Dataset Generation via Large Language Models | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
DATAGEN: UNIFIED SYNTHETIC DATASET VIA LARGE
LANGUAGE MODELS
Yue Huang1,†, Siyuan Wu2,†, Chujie Gao3, Dongping Chen2,4, Qihui Zhang5, Yao Wan2,∗
Tianyi Zhou6, Chaowei Xiao7, Jianfeng Gao8, Lichao Sun9,∗, Xiangliang Zhang1,∗
1University of Notre Dame, 2Huazhong University ... |
GR0y0F3Ipd | MAPS: Advancing Multi-Modal Reasoning in Expert-Level Physical Science | [
8,
6,
6,
6
] | Published as a conference paper at ICLR 2025
MAPS: ADVANCING MULTI-MODAL REASONING IN
EXPERT-LEVEL PHYSICAL SCIENCE
Erle Zhu1,2, Yadi Liu2, Zhe Zhang2, Xujun Li1,2, Jin Zhou2,
Xinjie Yu3, Minlie Huang1,2, Hongning Wang1,2,∗
1The Conversational AI (CoAI) Group, 2Department of Computer Science & Technology
3Department ... |
kGvXIlIVLM | Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment | [
6,
8,
6,
6,
8,
8
] | Published as a conference paper at ICLR 2025
TOWARD GUIDANCE-FREE AR VISUAL GENERATION
VIA CONDITION CONTRASTIVE ALIGNMENT
Huayu Chen1, Hang Su1, Peize Sun2, Jun Zhu1,3∗
1Department of Computer Science & Technology, Institute for AI, BNRist Center,
Tsinghua-Bosch Joint ML Center, THBI Lab, Tsinghua University
2The Un... |
oI5tZaWkF9 | Not All LLM-Generated Data Are Equal: Rethinking Data Weighting in Text Classification | [
6,
8,
8,
8
] | Published as a conference paper at ICLR 2025
NOT ALL LLM-GENERATED DATA ARE EQUAL:
RETHINKING DATA WEIGHTING IN TEXT CLASSIFICA-
TION
Hsun-Yu Kuo∗1,2,3,†, Yin-Hsiang Liao∗1,2,†, Yu-Chieh Chao1, Wei-Yun Ma1,†,‡, Pu-Jen Cheng2,†
1Academia Sinica 2National Taiwan University
3Swiss Federal Institute of Technology in Laus... |
Tn8EQIFIMQ | Language Models Trained to do Arithmetic Predict Human Risky and Intertemporal Choice | [
8,
6,
6,
8
] | Published as a conference paper at ICLR 2025
LANGUAGE MODELS TRAINED TO DO ARITHMETIC
PREDICT HUMAN RISKY AND INTERTEMPORAL
CHOICE
Jian-Qiao Zhu
Department of Computer Science
Princeton University
jz5204@princeton.edu
Thomas L. Griffiths
Department of Psychology and Computer Science
Princeton University
Haijiang Ya... |
wg1PCg3CUP | Scaling Laws for Precision | [
8,
8,
8,
8
] | Published as a conference paper at ICLR 2025
SCALING LAWS FOR PRECISION
Tanishq Kumar∗ 1 Zachary Ankner* 3,4 Benjamin F. Spector2 Blake Bordelon1
Niklas Muennighoff2 Mansheej Paul4 Cengiz Pehlevan1 Christopher R´e2
Aditi Raghunathan5
1Harvard University
4Databricks
3MIT
2Stanford University
5Carnegie Mellon Univers... |
599F4CZ0HB | Bench-O-Matic: Automating Benchmark Curation from Crowdsourced Data | [
5,
8,
5
] | Under review as a conference paper at ICLR 2025
BENCH-O-MATIC: AUTOMATING BENCHMARK
CURATION FROM CROWDSOURCED DATA
Anonymous authors
Paper under double-blind review
ABSTRACT
The rapid evolution of Large Language Models (LLMs) has outpaced the develop-
ment of model evaluation, highlighting the need for continuous ... |
E2PFv7ad3p | Have the VLMs Lost Confidence? A Study of Sycophancy in VLMs | [
8,
6,
6
] | Published as a conference paper at ICLR 2025
HAVE THE VISION-LANGUAGE MODELS LOST CONFI-
DENCE? A STUDY OF SYCOPHANCY IN VLMS
Shuo Li˚,1,Tao Ji˚,1, Xiaoran Fan˚,1
Linsheng Lu1, Leyi Yang1, Yuming Yang1, Zhiheng Xi1, Rui Zheng1
Yuran Wang2, Xiaohui Zhao2, Tao Gui:,1, Qi Zhang1, Xuanjing Huang1
Fudan University1 Honor ... |
WjKea8bGFF | Building Math Agents with Multi-Turn Iterative Preference Learning | [
6,
8,
8,
6
] | Published as a conference paper at ICLR 2025
BUILDING MATH AGENTS WITH MULTI-TURN ITERA-
TIVE PREFERENCE LEARNING
Wei Xiong 1, Chengshuai Shi2, Jiaming Shen3, Aviv Rosenberg4, Zhen Qin3, Daniele Calandriello3
Misha Khalman3, Rishabh Joshi3, Bilal Piot3, Mohammad Saleh3, Chi Jin5, Tong Zhang1, Tianqi Liu3
University o... |
DgaY5mDdmT | MLLMs Know Where to Look: Training-free Perception of Small Visual Details with Multimodal LLMs | [
5,
8,
8
] | Published as a conference paper at ICLR 2025
MLLMS KNOW WHERE TO LOOK:
TRAINING-FREE PERCEPTION OF
SMALL VISUAL DETAILS WITH MULTIMODAL LLMS
Jiarui Zhang , Mahyar Khayatkhoei
, Prateek Chhikara , Filip Ilievski
University of Southern California, USA
Vrije Universiteit Amsterdam, The Netherlands
ABSTRACT
Multimoda... |
fGIqGfmgkW | OpenPRM: Building Open-domain Process-based Reward Models with Preference Trees | [
8,
5,
5,
6
] | Published as a conference paper at ICLR 2025
OPENPRM: BUILDING OPEN-DOMAIN PROCESS-
BASED REWARD MODELS WITH PREFERENCE TREES
Kaiyan Zhang1
Xingtai Lv1 Ning Ding1 Biqing Qi4 Bowen Zhou1,4 ∗
Jiayuan Zhang2 Haoxin Li1 Xuekai Zhu3 Ermo Hua1
1 Tsinghua University
4 Shanghai Artificial Intelligence Laboratory
2 Beihang... |
mVCcWCjeEz | ToEdit: How to Synthesize Text Data to Avoid Model Collapse? | [
3,
8,
8,
6
] | Under review as a conference paper at ICLR 2025
TOEDIT: HOW TO SYNTHESIZE TEXT DATA TO AVOID
MODEL COLLAPSE?
Anonymous authors
Paper under double-blind review
ABSTRACT
We explore model collapse caused by synthetic data, where AI models trained on
such data experience a gradual decline in performance. Our initial an... |
zG459X3Xge | VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
VISRAG: VISION-BASED RETRIEVAL-AUGMENTED
GENERATION ON MULTI-MODALITY DOCUMENTS
Shi Yu1∗, Chaoyue Tang2∗, Bokai Xu2∗, Junbo Cui2∗, Junhao Ran3, Yukun Yan1†,
Zhenghao Liu4, Shuo Wang1, Xu Han1, Zhiyuan Liu1† , Maosong Sun1
1Department of Computer Science and Technology, Tsi... |
8KQzoD5XAr | CraftRTL: High-quality Synthetic Data Generation for Verilog Code Models with Correct-by-Construction Non-Textual Representations and Targeted Code Repair | [
8,
6,
6,
8
] | Published as a conference paper at ICLR 2025
CRAFTRTL: HIGH-QUALITY
SYNTHETIC DATA
GENERATION FOR VERILOG CODE MODELS WITH
CORRECT-BY-CONSTRUCTION NON-TEXTUAL REP-
RESENTATIONS AND TARGETED CODE REPAIR
Mingjie Liu∗, Yun-Da Tsai∗, Wenfei Zhou, Haoxing Ren
NVIDIA Corporation
{mingjiel, yundat, wenfeiz, haoxingr}@nvidia... |
w5ZtXOzMeJ | Auto-GDA: Automatic Domain Adaptation for Efficient Grounding Verification in Retrieval Augmented Generation | [
8,
6,
6
] | Published as a conference paper at ICLR 2025
AUTO-GDA: AUTOMATIC DOMAIN ADAPTATION
FOR GROUNDING VERIFICATION IN RETRIEVAL-
AUGMENTED GENERATION
Tobias Leemann∗
University of Tübingen
tobias.leemann@uni-tuebingen.de
Periklis Petridis∗
MIT
periklis@mit.edu
Giuseppe Vietri
AWS AI Labs
vietrigv@amazon.com
Dionysis Ma... |
1GTARJhxtq | Perplexed by Perplexity: Perplexity-Based Data Pruning With Small Reference Models | [
8,
6,
6,
6
] | Published as a conference paper at ICLR 2025
PERPLEXED BY PERPLEXITY: PERPLEXITY-BASED
DATA PRUNING WITH SMALL REFERENCE MODELS
Zachary Ankner 1,2 Cody Blakeney1 Kartik Sreenivasan1
Max Marion1 Matthew L. Leavitt3 Mansheej Paul1
1Databricks
2MIT 3DatologyAI
ABSTRACT
In this work, we investigate whether small langu... |
yaQbTAD2JJ | Language-Image Models with 3D Understanding | [
6,
6,
6,
6
] | Published as a conference paper at ICLR 2025
LANGUAGE-IMAGE MODELS WITH 3D
UNDERSTANDING
Jang Hyun Cho1,∗ Boris Ivanovic2 Yulong Cao2
Edward Schmerling2 Yue Wang2
Xinshuo Weng2
Boyi Li2
Yurong You2
Philipp Krähenbühl1,†
Yan Wang2,†
Marco Pavone2,†
1 University of Texas at Austin
2 NVIDIA
ABSTRACT
Multi-mod... |
Bgz3okeZ7H | AoPS Dataset: Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation | [
8,
8,
6,
3
] | Under review as a conference paper at ICLR 2025
AOPS DATASET: LEVERAGING ONLINE OLYMPIAD-
LEVEL MATH PROBLEMS FOR LLMS TRAINING AND
CONTAMINATION-RESISTANT EVALUATION
Anonymous authors
Paper under double-blind review
ABSTRACT
Advances in Large Language Models (LLMs) have sparked interest in their abil-
ity to solve... |
x1Bk51SCL9 | Face-Human-Bench: A Comprehensive Benchmark of Face and Human Understanding for Multi-modal Assistants | [
8,
3,
6,
6
] | Under review as a conference paper at ICLR 2025
FACE-HUMAN-BENCH: A COMPREHENSIVE
BENCHMARK OF FACE AND HUMAN UNDERSTANDING
FOR MULTI-MODAL ASSISTANTS
Anonymous authors
Paper under double-blind review
ABSTRACT
Faces and humans are crucial elements in social interaction and are widely in-
cluded in everyday photos a... |
txoJvjfI9w | PEARL: Towards Permutation-Resilient LLMs | [
6,
8,
8,
3
] | Published as a conference paper at ICLR 2025
PEARL: TOWARDS PERMUTATION-RESILIENT LLMS
Liang Chen1 Li Shen2∗ Yang Deng3 Xiaoyan Zhao1 Bin Liang1 Kam-Fai Wong1∗
1The Chinese University of Hong Kong 2Shenzhen Campus of Sun Yat-sen University 3SMU
{lchen, kfwong}@se.cuhk.edu.hk mathshenli@gmail.com
ABSTRACT
The in-con... |
lja4JMesmC | From Generalist to Specialist: Adapting Vision Language Models via Task-Specific Visual Instruction Tuning | [
5,
6,
8,
6
] | Under review as a conference paper at ICLR 2025
FROM GENERALIST TO SPECIALIST: ADAPTING VI-
SION LANGUAGE MODELS VIA TASK-SPECIFIC VI-
SUAL INSTRUCTION TUNING
Anonymous authors
Paper under double-blind review
ABSTRACT
Large vision language models (VLMs) combine large language models with vi-
sion encoders, demonstr... |
tZCqSVncRf | MIRAGE: Evaluating and Explaining Inductive Reasoning Process in Language Models | [
5,
6,
8,
5
] | Published as a conference paper at ICLR 2025
MIRAGE: EVALUATING AND EXPLAINING INDUCTIVE
REASONING PROCESS IN LANGUAGE MODELS
Jiachun Li1,2, Pengfei Cao1,2, Zhuoran Jin1,2, Yubo Chen1,2,∗, Kang Liu1,2, Jun Zhao1,2
1School of Artificial Intelligence, University of Chinese Academy of Sciences
2The Key Laboratory of Cog... |
XrsOu4KgDE | Attributing Culture-Conditioned Generations to Pretraining Corpora | [
5,
6,
8,
8,
8
] | Published as a conference paper at ICLR 2025
ATTRIBUTING CULTURE-CONDITIONED
GENERATIONS TO PRETRAINING CORPORA
Huihan Li1∗ Arnav Goel2∗ Keyu He1 Xiang Ren1
1University of Southern California
2IIIT Delhi
{huihanl,frankhe,xiangren}@usc.edu,arnav21519@iiitd.ac.in
ABSTRACT
In open-ended generative tasks like narrative... |
e9yfCY7Q3U | Improved Techniques for Optimization-Based Jailbreaking on Large Language Models | [
6,
8,
6,
5
] | Published as a conference paper at ICLR 2025
IMPROVED TECHNIQUES FOR OPTIMIZATION-BASED
JAILBREAKING ON LARGE LANGUAGE MODELS
Xiaojun Jia1,2, Tianyu Pang†2, Chao Du2, Yihao Huang1, Jindong Gu3, Yang Liu1,
Xiaochun Cao†4, Min Lin2
1Nanyang Technological University, Singapore
2Sea AI Lab, Singapore
3University of Oxfor... |
oqsQbn4XfT | On the Diversity of Synthetic Data and its Impact on Training Large Language Models | [
8,
6,
6,
3,
6
] | Under review as a conference paper at ICLR 2025
ON THE DIVERSITY OF SYNTHETIC DATA AND ITS IM-
PACT ON TRAINING LARGE LANGUAGE MODELS
Anonymous authors
Paper under double-blind review
ABSTRACT
The rise of Large Language Models (LLMs) has accentuated the need for diverse,
high-quality pre-training data. Synthetic da... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.