AI Paper Atlas

Research map / Continual learning / Continual fine-tuning of LLMs

COMET: Decoupled Distillation, Routing, and Capacity Control for Task-Agnostic Continual Vision--Language Learning

Seungyoon Woo, Keonhee Park, Gunhee Kim

NeurIPS 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Mixture-of-experts for multimodal continual tuning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes COMET, a task-agnostic continual learning method for vision-language models (e.g., CLIP) that decouples distillation, routing, and capacity control to address modal-consistency drift and related pressures.

Problem

Task-agnostic continual learning for VLMs such as CLIP, where tasks arrive sequentially without past data or task identity and the backbone cannot be retrained, causing modal-consistency drift and representation-related interference beyond the usual stability-plasticity tradeoff.

Approach

Decouples three mechanisms (distillation, routing, and capacity control) to jointly handle the coupled pressures of task-agnostic continual vision-language learning; details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Tri-Affinity Distillation (TAD) (ablation)COMET base model · X-TAIL full-shotTransfer: 56.6 -> 61.3Improved
add Student-Aware Cross-Expert Distillation (SACED) alone (ablation)COMET base model · X-TAIL full-shotTransfer: 56.6 -> 59.9Improved
add SACED alone (ablation)COMET base model · X-TAIL full-shotLast: 85.6 -> 85.3No clear effect
add SACED on top of TAD (ablation)COMET · X-TAIL full-shotTransfer / Average / Last: 61.3 -> 62.6 Transfer; 75.4 -> 76.1 Average; 87.6 -> 88.1 LastImproved
add both TAD and SACED in 16-shot setting (ablation)COMET · X-TAIL 16-shotLast: 73.3 -> 83.3Improved
use bandit-controlled merging instead of Always-Expand (ablation)COMET · X-TAIL full-shotTransfer: 61.6 -> 62.6 with 10 -> 8 expertsImproved
use Always-Merge or random merging instead of CMAB (ablation)COMET · X-TAIL full-shotLast: 88.1 -> 58.4 (Always-Merge); 69.1 (Random)Hurt
add SACED to reduce routing-induced interference (ablation)COMET · X-TAIL misrouted samplesaccuracy on misrouted samples: +10.8 pp; 52.5% of misrouted still correctImproved
use full per-task experts with bandit merging instead of adaptersCOMET · heterogeneous 4-taskTransfer: 12.5 -> 31.2Improved
use smaller TinyCLIP per-task experts (58M) instead of CLIP-B/16 (150M)COMET · X-TAIL full-shotTransfer / Average / Last: vs LADA: +1.5, +1.4, +1.4 ppImproved
use reconstruction router with sparse top-1 pathCOMET · inference costlatency (ms): 176.9 -> 107.1 (vs LADA)Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy