ARXIV:2604.25898 · CONTINUAL RL · SUBMITTED 29 APR · 03:16 UTC · FRESHNESS STALE

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning

Dominik Żurek · Kamil Faber · Marcin Pietron · Paweł Gajewski · Roberto Corizzo · arXiv

TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.

Ship in 2-4 weeks›Score7.0Evidence unverified

Opportunity summary

Pain TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.

Evidence 0 refs | 4 sources | 67% coverage

Blocker Evidence unverified

Open Build Read PDF Signal Canvas Track

PROBLEM

TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance. This setting corresponds to domains where new tasks arise over time, but adapting…

METHOD

Full abstract

Continual offline reinforcement learning (CORL) aims to learn a sequence of tasks from datasets collected over time while preserving performance on previously learned tasks. This setting corresponds to domains where new tasks arise over time, but adapting the model in live environment interactions is expensive, risky, or impossible. However, CORL inherits the dual difficulty of offline reinforcement learning and adapting while preventing catastrophic forgetting. Replay-based continual learning approaches remain a strong baseline but incur memory overhead and suffer from a distribution mismatch between replayed samples and newly learned policies. At the same time, architectural continual learning methods have shown strong potential in supervised learning but remain underexplored in CORL. In this work, we propose TSN-Affinity, a novel CORL method based on TinySubNetworks and Decision Transformer. The method enables task-specific parameterization and controlled knowledge sharing through a RL-aware reuse strategy that routes tasks according to action compatibility and latent similarity. We evaluate the approach on benchmarks based on Atari games and simulations of manipulation tasks with the Franka Emika Panda robotic arm, covering both discrete and continuous control. Results show strong retention from sparse SubNetworks, with routing further improving multi-task performance. Our findings suggest that similarity-guided architectural reuse is a strong and viable alternative to replay-based strategies in a CORL setting. Our code is available at: https://github.com/anonymized-for-submission123/tsn-affinity.

RESULT

ScienceToStartup currently rates this 7.0/10 on the public viability pass. The method enables task-specific parameterization and controlled knowledge sharing through a RL-aware reuse strategy that routes tasks according to action compatibility and latent similarity.…

WHY NOW

Continual RL moved forward this cycle; last verified April 2026. Public score 7.0/10. Implementation evidence is present through a linked repository.

Continue into Read for claims, analysis, references, and neighboring papers.

Opportunity summary

Score7.0

PainTSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.

Evidence0 refs | 4 sources | 67% coverage

Blockerno shell-level blocker reported

Analysis summary

TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

Competitive landscape

TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.

Segment

Continual RL

Adoption evidence

Public code linked for build inspection

Commercial read

7.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

{ "contract_version": "paper-r2", "paper_id": "43f0a57d-b08a-457b-a409-04c9905ef812", "arxiv_id": "2604.25898", "canonical_route": "/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning", "active_tab": "synced from current hash by the drawer client", "selected_artifact": "tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning", "endpoints": { "paper_pack": "/api/v1/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning/paper-pack", "build_passport": "/api/v1/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning/build-passport", "mcp_resource": "sciencetostartup://surfaces/paper-workspace" } }

{ "surface": "paper", "mode": "paper", "query": "TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning", "normalized_query": "2604.25898", "route": "/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning", "paper_ref": "tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning", "topic_slug": null, "benchmark_ref": null, "dataset_ref": null }

{ "@context": "https://schema.org", "@graph": [ { "@type": "WebPage", "@id": "https://sciencetostartup.com/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning#webpage", "url": "https://sciencetostartup.com/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning", "name": "TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning", "description": "TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.", "isPartOf": { "@id": "https://sciencetostartup.com/#website" } }, { "@type": "ScholarlyArticle", "@id": "https://sciencetostartup.com/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning#scholarlyArticle", "headline": "TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning", "description": "TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.", "url": "https://sciencetostartup.com/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning", "sameAs": "https://arxiv.org/abs/2604.25898", "identifier": { "@type": "PropertyValue", "propertyID": "arXiv", "value": "2604.25898" }, "isAccessibleForFree": true, "isPartOf": { "@id": "https://sciencetostartup.com/#website" }, "datePublished": "2026-04-28T17:41:04.000Z", "author": [ { "@type": "Person", "name": "Dominik Żurek" }, { "@type": "Person", "name": "Kamil Faber" }, { "@type": "Person", "name": "Marcin Pietron" }, { "@type": "Person", "name": "Paweł Gajewski" }, { "@type": "Person", "name": "Roberto Corizzo" } ], "codeRepository": "https://github.com/anonymized-for-submission123/tsn-affinity", "additionalProperty": [ { "@type": "PropertyValue", "propertyID": "viabilityScore", "value": 7 }, { "@type": "PropertyValue", "propertyID": "researchDomain", "value": "Continual RL" }, { "@type": "PropertyValue", "propertyID": "commercialReadiness", "value": "code, repo url" } ] }, { "@type": "SoftwareSourceCode", "@id": "https://sciencetostartup.com/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning#software", "name": "TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning - Source Code", "description": "TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.", "codeRepository": "https://github.com/anonymized-for-submission123/tsn-affinity", "url": "https://github.com/anonymized-for-submission123/tsn-affinity" }, { "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://sciencetostartup.com" }, { "@type": "ListItem", "position": 2, "name": "Continual RL", "item": "https://sciencetostartup.com/topics" }, { "@type": "ListItem", "position": 3, "name": "TSN-Affinity: Similarity-Driven Parameter Reuse for Continua", "item": "https://sciencetostartup.com/paper/tsn-affinity-similarity-driven-parameter-reuse-for-continual-offline-reinforcement-learning" } ] } ] }

Competitive landscape

TSN-Affinity is a continual offline reinforcement learning method that uses similarity-driven parameter reuse to prevent catastrophic forgetting and improve multi-task performance.

Segment

Continual RL

Adoption evidence

Public code linked for build inspection

Commercial read

7.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning

TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement Learning

Claim map

Constellation map

Competitive landscape

Buzz

PDF

REFERENCES

Related Papers

Subscribe to the weekly brief

Build artifacts

Brief

Experiment plan

Validation checklist

Scientific founder

Translational engineer

Domain operator

GTM lead

Regulatory/clinical advisor

Timeline

Claim map

Constellation map

Competitive landscape

Buzz

PDF

REFERENCES

Related Papers

Subscribe to the weekly brief

Build artifacts

Brief

Experiment plan

Validation checklist

Scientific founder

Translational engineer

Domain operator

GTM lead

Regulatory/clinical advisor

Timeline