Decomposing the Basic Abilities of Large Language Models: Mitigating Cross-Task Interference in Multi-Task Instruct-Tuning

• URL: http://arxiv.org/abs/2605.05676v1
• Date: Thu, 07 May 2026 05:08:58 GMT
• Title: Decomposing the Basic Abilities of Large Language Models: Mitigating Cross-Task Interference in Multi-Task Instruct-Tuning
• Abstract summary: Multi-task instruct-tuning in large language models (LLMs) has been largely driven by multi-task instruct-tuning.<n>The training paradigm suffers from a key issue, named cross-task interference, due to conflicting gradients over shared parameters among different tasks.<n>We propose BADIT that decomposes LLM parameters into high-singular-value LoRA experts representing basic abilities.<n>We conduct extensive experiments on the SuperNI benchmark with 6 LLMs, and empirical results demonstrate that BADIT can outperform SOTA methods and mitigate the degree of cross-task interference.
• Score: 58.90602461025722
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recently, the prominent performance of large language models (LLMs) has been largely driven by multi-task instruct-tuning. Unfortunately, this training paradigm suffers from a key issue, named cross-task interference, due to conflicting gradients over shared parameters among different tasks. Some previous methods mitigate this issue by isolating task-specific parameters, e.g., task-specific neuron selection and mixture-of-experts. In this paper, we empirically reveal that the cross-task interference still exists for the existing solutions because of many parameters also shared by different tasks, and accordingly, we propose a novel solution, namely Basic Abilities Decomposition for multi-task Instruct-Tuning (BADIT). Specifically, we empirically find that certain parameters are consistently co-activated, and that co-activated parameters naturally organize into base groups. This motivates us to analogize that LLMs encode several orthogonal basic abilities, and that any task can be represented as a linear combination of these abilities. Accordingly, we propose BADIT that decomposes LLM parameters into orthogonal high-singular-value LoRA experts representing basic abilities, and dynamically enforces their orthogonality during training via spherical clustering of rank-1 components. We conduct extensive experiments on the SuperNI benchmark with 6 LLMs, and empirical results demonstrate that BADIT can outperform SOTA methods and mitigate the degree of cross-task interference.

Related papers

• Essential Subspace Merging for Multi-Task Learning [78.5390284258307]
  We analyze the output shifts induced by task updates and observe that their energy is concentrated in a small number of principal directions.<n>Most remaining directions carry little task-relevant energy, but their accumulation across multiple task updates can cause severe interference during merging.<n>Based on ESD, we introduce Essential Subspace Merging (ESM), a training-free static merging method.<n>We extend ESM to ESM++, a training-free dynamic merging method that decomposes task-specific residuals into low-rank experts.
  arXiv  Detail & Related papers  (2026-06-17T15:07:11Z)
• Model Merging in the Essential Subspace [78.5390284258307]
  Model merging aims to integrate multiple task-specific fine-tuned models into a single multi-task model without additional training.<n>Despite extensive research, task interference remains a major obstacle that often undermines the performance of merged models.<n>We propose ESM (Essential Subspace Merging), a robust framework for effective model merging.
  arXiv  Detail & Related papers  (2026-02-23T00:33:38Z)
• DPI: Exploiting Parameter Heterogeneity for Interference-Free Fine-Tuning [11.751530422766836]
  Supervised fine-tuning (SFT) is a crucial step for adapting large language models (LLMs) to downstream tasks.<n>We propose a principled approach to disentangle and isolate task-specific parameter regions.
  arXiv  Detail & Related papers  (2026-01-25T10:30:45Z)
• Bigger, Regularized, Categorical: High-Capacity Value Functions are Efficient Multi-Task Learners [60.75160178669076]
  We show that the use of high-capacity value models trained via cross-entropy and conditioned on learnable task embeddings addresses the problem of task interference in online reinforcement learning.<n>We test our approach on 7 multi-task benchmarks with over 280 unique tasks, spanning high degree-of-freedom humanoid control and discrete vision-based RL.
  arXiv  Detail & Related papers  (2025-05-29T06:41:45Z)
• MoSLD: An Extremely Parameter-Efficient Mixture-of-Shared LoRAs for Multi-Task Learning [8.868481107848185]
  MoSLD is a mixture-of-shared-LoRAs model with a dropout strategy.<n>MoSLD addresses challenges by sharing the upper projection matrix in LoRA among different experts.<n>Our model exhibits excellent performance in both single-task and multi-task scenarios.
  arXiv  Detail & Related papers  (2024-12-12T05:22:49Z)
• MoDE: Effective Multi-task Parameter Efficient Fine-Tuning with a Mixture of Dyadic Experts [6.245113492272563]
  Mixture of Dyadic Experts (MoDE) is a novel design for efficient multi-task adaptation. Our design allows for more fine-grained mixing, thereby increasing the model's ability to jointly handle multiple tasks.
  arXiv  Detail & Related papers  (2024-08-02T18:05:10Z)
• Merging Multi-Task Models via Weight-Ensembling Mixture of Experts [64.94129594112557]
  Merging Transformer-based models trained on different tasks into a single unified model can execute all the tasks concurrently. Previous methods, exemplified by task arithmetic, have been proven to be both effective and scalable. We propose to merge most of the parameters while upscaling the Transformer layers to a weight-ensembling mixture of experts (MoE) module.
  arXiv  Detail & Related papers  (2024-02-01T08:58:57Z)
• Concrete Subspace Learning based Interference Elimination for Multi-task Model Fusion [86.6191592951269]
  Merging models fine-tuned from common extensively pretrained large model but specialized for different tasks has been demonstrated as a cheap and scalable strategy to construct a multitask model that performs well across diverse tasks. We propose the CONtinuous relaxation dis (Concrete) subspace learning method to identify a common lowdimensional subspace and utilize its shared information track interference problem without sacrificing performance.
  arXiv  Detail & Related papers  (2023-12-11T07:24:54Z)
• Mitigating Task Interference in Multi-Task Learning via Explicit Task Routing with Non-Learnable Primitives [19.90788777476128]
  Multi-task learning (MTL) seeks to learn a single model to accomplish multiple tasks by leveraging shared information among the tasks. Existing MTL models have been known to suffer from negative interference among tasks. We propose ETR-NLP to mitigate task interference through a synergistic combination of non-learnable primitives and explicit task routing.
  arXiv  Detail & Related papers  (2023-08-03T22:34:16Z)
• Reparameterizing Convolutions for Incremental Multi-Task Learning without Task Interference [75.95287293847697]
  Two common challenges in developing multi-task models are often overlooked in literature. First, enabling the model to be inherently incremental, continuously incorporating information from new tasks without forgetting the previously learned ones (incremental learning) Second, eliminating adverse interactions amongst tasks, which has been shown to significantly degrade the single-task performance in a multi-task setup (task interference)
  arXiv  Detail & Related papers  (2020-07-24T14:44:46Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.