@TmlrSub profile picture
New Submissions to TMLR @TmlrSub

Submissions to Transactions of Machine Learning Research

Joined March 2022
  • Tweets 5,718
  • Following 3
  • Followers 2,337
532 Photos and videos
POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with Prompt-Optimized Parameter Shaking openreview.net/forum?id=wMiE… #adversarial #multimodal #malicious

POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with...

Multimodal Large Language Models (MLLMs) have demonstrated impressive performance on cross-modal tasks by jointly training on large-scale textual and visual data, where privacy-sensitive examples...

openreview.net
2
110
SSM-PixNav: State Space Models for Pixel-Guided Embodied Navigation openreview.net/forum?id=RmsM… #navigating #navigation #robot

SSM-PixNav: State Space Models for Pixel-Guided Embodied Navigation

While navigating a robot towards a specific object or image, how do we ensure it focuses on the correct location we are referring to? Object goal navigation, image goal navigation, goal instance...

openreview.net
1
143
The Impact of Enforcing Representational Consistency of Identical Transformations for Disentangled Representation openreview.net/forum?id=VjbB… #autoencoders #representational #regularization

The Impact of Enforcing Representational Consistency of Identical...

Recent symmetry-based approaches in Variational Autoencoders (VAEs) have advanced disentanglement learning and compositional generalization. However, existing methods can encode identical semantic...

openreview.net
120
Inducing Artificial Uncertainty in Language Models openreview.net/forum?id=cUcW… #supervised #uncertainty #datasets

Inducing Artificial Uncertainty in Language Models

In safety-critical applications, language models should be able to characterize their uncertainty with meaningful probabilities. Many uncertainty quantification approaches require supervised data;...

openreview.net
86
Observer-Side Diagnosis of Prompt-Induced Interference in Large Language Models: A Macro-Group Vocabulary and Targeted Cross-Lingual Stress Tests openreview.net/forum?id=zcGz… #interpretability #politeness #prompt

Observer-Side Diagnosis of Prompt-Induced Behavioral Shifts in...

Small prompt fragments can change not only an intended surface property of a response, such as register, but also observable markers of uncertainty, conditionality, and scope. This paper introduces...

openreview.net
90
Recursive Deep Inverse Reinforcement Learning openreview.net/forum?id=bYIt… #adversarial #adversary #learning

Recursive Deep Inverse Reinforcement Learning

Inferring an adversary’s goals from exhibited behavior is crucial for counterplanning and non-cooperative multi-agent systems in domains like cybersecurity, military, and strategy games. Deep...

openreview.net
74
Client-Level Defense Placement for Adversarially Robust Federated Reinforcement Learning openreview.net/forum?id=JRpL… #defenses #adversarially #attackers

Client-Level Defense Placement for Adversarially Robust Federated...

Federated Reinforcement Learning (FRL) extends federated learning to sequential decision-making, enabling multiple clients to collaboratively train a global policy without sharing raw trajectories....

openreview.net
68
Monte Carlo Multi-Feature Baseline Shapley (MMBS): An axiomatic attribution method for fine-grained explanations of image classification networks openreview.net/forum?id=LLFI… #features #feature #shapley

Monte Carlo Multi-Feature Baseline Shapley (MMBS): An axiomatic...

This paper presents the Multi-Feature Baseline Shapley (MBS) attribution method for explaining the outcome of a neural network for a given input. MBS generalizes the Integrated Gradients (IG) and...

openreview.net
117
ViP$^2$-CLIP: Visual-Perception Prompting with Unified Alignment for Zero-Shot Anomaly Detection openreview.net/forum?id=KCRR… #anomaly #visual #anomalies

ViP$^2$-CLIP: Visual-Perception Prompting with Unified Alignment...

Zero-Shot Anomaly Detection (ZSAD) aims to detect anomalies in a target dataset without any training samples, leveraging models trained on auxiliary data. While CLIP offers strong cross-modal...

openreview.net
89
Verify What Matters: Budgeted Verification for Tool Using Agents under Counterfactual Downstream Harm openreview.net/forum?id=nv1j… #failures #counterfactual #uncertainty

Verify What Matters: Budgeted Verification for Tool Using Agents...

Tool-using agents make intermediate decisions that alter persistent state, shape later observations, and can trigger failures that are not equally easy to recover from. When verification is costly,...

openreview.net
1
80
Cog-VADU: A Training-Free Cognitive Reasoning Framework for Video Anomaly Detection and Understanding openreview.net/forum?id=QcuS… #videos #visual #recurrent

Cog-VADU: A Training-Free Cognitive Reasoning Framework for Video...

Video Anomaly Detection (VAD) aims to temporally localize abnormal events in videos. Most existing approaches rely on dataset-specific training and curated annotations, limiting generalization in...

openreview.net
82
On the Expressive Power and Limitations of Multi-Layer SSMs openreview.net/forum?id=NQ8I… #expressiveness #expressive #ssms

On the Expressive Power and Limitations of Multi-Layer SSMs

We study the expressive power and limitations of multi-layer state-space models (SSMs). First, we show that multi-layer SSMs face fundamental limitations in compositional tasks, revealing an...

openreview.net
80
Can LLMs Rank Candidates with Missing Sensitive Attributes Fairly? openreview.net/forum?id=VrAs… #fairness #attributes #ranking

Can LLMs Rank Candidates with Missing Sensitive Attributes Fairly?

Large language models (LLMs) are increasingly deployed in high-stakes ranking systems used for hiring, lending, and scholarship allocation, raising concerns about fairness, accountability, and...

openreview.net
62
Adversarial Attacks and Defenses in Vision-Language Pre- training: Techniques, Challenges and Opportunities openreview.net/forum?id=I2zO… #adversarial #vulnerabilities #security

Adversarial Attacks and Defenses in Vision-Language Pre- training:...

Vision-language pretraining (VLP) has emerged as a powerful paradigm for multimodal learning. However, despite their superior capabilities, VLPs remain vulnerable to adversarial attacks by...

openreview.net
1
81
HyperDG: Hyperbolic Representation Alignment for Robust Domain Generalization via Curvature Refinement openreview.net/forum?id=TSsh… #manifold #manifolds #generalization

HyperDG: Hyperbolic Representation Alignment for Robust Domain...

Domain generalization often suffers from geometric inconsistencies in representations learned across multiple source domains. Although recent approaches pursue flat minima or invariant features...

openreview.net
1
99
Learning from Complaints: Adversarial Disentanglement for Robust Scalper Detection in E-Commerce Promotions openreview.net/forum?id=yrYe… #adversarial #complaints #complaint

Learning from Complaints: Adversarial Disentanglement for Robust...

Identifying scalpers in e-commerce promotions is a critical challenge where instance- dependent label noise is pervasive: legitimate users with ambiguous patterns (e.g., frequent on-the-hour...

openreview.net
84
Variance Reduction in Sketching Algorithms via Complex Random Variables openreview.net/forum?id=0Czp… #rademacher #countsketch #tensorsketch

Variance Reduction in Sketching Algorithms via Complex Random...

The seminal \texttt{CountSketch} algorithm of~\cite{count_sketch} compresses high-dimensional real-valued vectors while approximately preserving pairwise inner products in time proportional to the...

openreview.net
1
97
Investigating the limits of free-form debate as a scalable oversight strategy openreview.net/forum?id=vRCG… #debaters #debater #debate

Investigating the limits of free-form debate as a scalable...

Debate is a scalable oversight method involving two copies of a strong model trained to defend alternative responses to a question, with a judge with less task-relevant information, time, or...

openreview.net
1
128
Enhancing Self-Supervised Visual Representation Learning via Low-Rank Adapted LLMs openreview.net/forum?id=s2T8… #visual #enhance #enhancer

Enhancing Self-Supervised Visual Representation Learning via...

The integration of Large Language Model (LLMs) blocks with Vision Transformers (ViTs) holds significant promise for vision-only tasks by leveraging the rich semantic knowledge and reasoning...

openreview.net
2
122
Do Object Channels Improve Robustness in Deep Reinforcement Learning? openreview.net/forum?id=7BFb… #pixels #pixel #atari

Do Object Channels Improve Robustness in Deep Reinforcement Learning?

Pixel-based reinforcement learning agents often exploit spurious visual correlations, leading to brittle policies that fail under minor visual perturbations. We systematically investigate spatial...

openreview.net
83
Load more