RAN1 /
Meeting 118bis /
NR_AIML_air
AI/ML for NR Air Interface
NR_AIML_air · 29 contributions
Sub-topics
6 agenda items · grouped by 3GPP agenda numbering
5
Incoming Liaison Statements
6 contributions
This sub-topic covers incoming liaison statements between RAN1 and RAN2 regarding AI/ML functionality reporting procedures for beam management UE-sided models, UE-side data collection capabilities, and cross-WG coordination on AI/ML framework development. Companies are providing responses to technical questions about signaling procedures, granularity levels, and network control mechanisms for AI/ML implementations in NR air interface.
Company positions
- Ericsson — Advocates for inclusion of UE-side data collection capabilities for UE-side model training in AI/ML applications, supporting RAN2's endorsement and requesting incorporation into technical report TR 38.843.
- Intel — Seeks RAN1's technical input and validation on agreed signaling procedures for AI/ML beam management functionality reporting, advocating for a structured 5-step process while acknowledging multiple aspects remain for further study.
- InterDigital — Advocates for a comprehensive framework balancing MNO control with standardized AI/ML data collection, pushing for solutions ensuring full MNO controllability and visibility while maintaining privacy compliance.
- Spreadtrum — Advocates for mandatory network-side additional conditions and structured configuration frameworks with sub-use case level granularity, opposing optional network-side conditions and functionality-specific approaches.
- vivo — Advocates for mandatory associated IDs as essential for AI/ML training-inference consistency and FG-specific granularity in functionality definition, while opposing providing inference configuration in early steps.
- ZTE — Advocates for flexible and optional network-side signaling to minimize overhead while supporting sub-use-case level granularity and reusing existing CSI framework mechanisms, opposing mandatory associated ID signaling.
Open issues
- Granularity level for AI/ML functionality reporting (sub-use case vs FG-specific vs functionality-specific)
- Whether associated IDs should be mandatory or optional for AI/ML training-inference consistency
- Network-side signaling requirements and overhead optimization for AI/ML beam management
- MNO controllability and visibility requirements for UE-side AI/ML data collection
- Technical feasibility of different AI/ML data collection options requiring SA group input
9.1.1
Specification support for beam management
7 contributions
Companies are discussing specification support for AI/ML-enabled beam management in NR Release 19, focusing on three key areas: data collection mechanisms for model training, inference procedures for beam prediction, and performance monitoring frameworks. The technical debate centers around how much to reuse existing CSI frameworks versus introducing new AI/ML-specific signaling, with companies proposing various approaches for both network-sided and UE-sided model implementations.
Company positions
- China Telecom — Advocates for a comprehensive AI/ML beam management framework supporting both network-sided and UE-sided models with flexible data collection mechanisms and beam prediction accuracy KPIs as the primary performance monitoring metric.
- FUTUREWEI — Strongly advocates for reusing existing CSI frameworks to minimize specification complexity and opposes introducing new complex metrics like probability information and confidence levels in favor of simpler, more testable approaches.
- Huawei — Advocates for pragmatic reuse of existing CSI framework and signaling mechanisms while opposing overly complex new procedures, and strongly supports expanding beam measurement capabilities up to 256 beams with unified solutions across BM-Case 1/2.
- Kyocera — Advocates for a comprehensive AI/ML beam management framework leveraging existing CSI infrastructure with minimal new specifications, strongly supporting virtual Set A configuration for UE-side models and flexible Set B definition through new information elements.
- Spreadtrum — Advocates for UE-initiated data collection for UE-side models, supports up to 16 beams reporting per instance, and favors reusing existing CSI frameworks to minimize specification impact while opposing larger quantization steps to maintain accuracy.
- Tejas Network Limited — Advocates for simplified beam management through single resource set configurations where possible, differential L1-RSRP reporting to reduce overhead, and flexible Associated ID mechanisms supporting both CSI framework and higher-layer signaling.
- ZTE — Strongly advocates for functionality-based lifecycle management over model-ID-based approaches and pushes for bitmap-based beam reporting methods to significantly reduce signaling overhead while leveraging existing UE capability frameworks.
Open issues
- Whether to introduce new AI/ML-specific signaling or maximize reuse of existing CSI frameworks
- How to balance beam measurement capability expansion (up to 256 beams) with signaling overhead concerns
- Choice between functionality-based versus model-ID-based lifecycle management approaches
- Optimal beam reporting methods and quantization steps for maintaining accuracy while reducing overhead
9.1.2
Specification support for positioning accuracy enhancement
4 contributions
This sub-topic focuses on AI/ML techniques to enhance positioning accuracy in NR air interface, with companies discussing model input/output specifications, measurement frameworks, and training procedures. Key technical debates center around sample-based versus path-based measurements, the inclusion of phase information in channel measurements, and whether to develop new signaling mechanisms or reuse existing 3GPP procedures.
Company positions
- Ericsson — Advocates for sample-based measurements over legacy path-based measurements for superior performance and lower complexity, while strongly opposing phase information inclusion in model inputs due to deployment costs and minimal accuracy gains.
- Huawei — Proposes pragmatic implementation-based solutions over rigid standardization, arguing that vendor consistency can resolve ambiguity issues without complex new signaling mechanisms and favoring reuse of legacy procedures.
- Tejas Network Limited — Advocates for leveraging existing Release-17 path-based measurement frameworks while emphasizing the need to address receiver implementation dependencies, opposing overly complex new measurement frameworks in favor of existing procedure reuse.
- ZTE — Strongly supports maximizing reuse of existing 3GPP procedures and advocates for CIR with phase information over PDP for better positioning accuracy despite higher overhead, while also favoring sample-based measurements over path-based approaches.
Open issues
- Whether to use sample-based or path-based measurement frameworks for AI/ML positioning enhancement
- Whether phase information should be included in channel measurements despite complexity and cost concerns
- How to balance standardization rigor versus implementation flexibility for AI/ML model specifications
- Whether to develop new signaling mechanisms or rely on existing 3GPP procedures for AI/ML positioning features
9.1.3
Specification support for CSI prediction
3 contributions
Companies are discussing specification support for CSI prediction in AI/ML-enhanced NR air interfaces, specifically focusing on how to ensure consistency between training and inference phases. The main debate centers around whether to introduce associated IDs and network-side mechanisms versus relying on UE-side performance monitoring approaches.
Company positions
- Huawei — Strongly advocates against introducing associated IDs or network-side indications for CSI prediction consistency, arguing they are unnecessary and create privacy risks. Proposes UE-side performance monitoring as a sufficient implementation-based solution instead.
- Spreadtrum — Advocates for reusing existing AI beam management conclusions and associated ID mechanisms for CSI prediction to minimize workload. Opposes performance monitoring-based approaches, arguing they would cause significant performance loss due to trial-and-error processes.
- ZTE — Proposes a systematic approach that prioritizes identifying additional conditions before developing detailed consistency solutions. Supports reusing proven AI beam prediction mechanisms but opposes introducing new associated IDs for scenarios/carrier frequency.
Open issues
- Whether associated IDs are necessary for ensuring training-inference consistency in CSI prediction
- Whether UE-side performance monitoring is sufficient versus requiring network-side mechanisms
- How to balance specification complexity with implementation flexibility for CSI prediction consistency
9.1.4.1
CSI compression
4 contributions
Companies are discussing AI/ML-based CSI compression for NR air interface, focusing on inter-vendor training collaboration approaches and temporal domain extensions. The main technical debate centers around different collaboration options including dataset sharing, model parameter exchange, and standardized reference architectures.
Company positions
- FUTUREWEI — Advocates for using Rel-16 eType II codebook with enhanced parameters for data collection and monitoring, and pushes for comprehensive analysis of inter-vendor collaboration options despite additional overhead concerns.
- Huawei — Supports Direction A (dataset/model sharing for UE-side offline engineering) over Direction B, and prioritizes Option 4 dataset sharing with simpler sub-options due to significantly lower air-interface overhead.
- Spreadtrum — Strongly advocates for AI-based CSI Spatial-Temporal-Frequency compression showing superior performance gains, while opposing option 5a for inter-vendor collaboration due to increased complexity.
- ZTE — Advocates for prioritizing Option 3 (standardized reference model structure with parameter exchange) using NW-first training and over-the-air delivery, while opposing Option 4 due to dataset exchange overhead concerns.
Open issues
- Choice between different inter-vendor collaboration approaches (Options 3, 4, 5a) with conflicting overhead and complexity trade-offs
- Whether to prioritize dataset sharing versus parameter exchange mechanisms
- Integration of temporal domain extensions with spatial-frequency compression methods
9.1.4.2
Other aspects of AI/ML model and data
5 contributions
Companies are discussing AI/ML model identification approaches, model transfer/delivery mechanisms, and UE-side training data collection for NR air interface. The debate centers around different model identification options (MI-Options 1-4), network vs UE-controlled ID assignment, and prioritization of various model transfer cases with consideration of cross-vendor collaboration complexity.
Company positions
- FUTUREWEI — Advocates for simplified model identification approaches with network-controlled model ID assignment and cell-scoped associated IDs, while deprioritizing complex offline cross-vendor collaboration mechanisms.
- Huawei — Proposes focusing model identification discussions exclusively on two-sided models while eliminating MI-Option 1, and advocates for Case y as baseline for UE-side training while opposing Case z1 due to cross-vendor collaboration issues.
- Spreadtrum — Advocates for deprioritizing model transfer cases z1 and z2 in Rel-19 due to cross-vendor collaboration burdens, supports mechanism 1a for UE data collection to avoid privacy exposure, and promotes network-controlled model identification.
- Tejas Network Limited — Advocates for UE-driven model ID assignment and flexible many-to-one mapping between associated IDs and model IDs to enable generalized models, while supporting current model transfer/delivery case prioritizations.
- ZTE — Strongly opposes MI-Option 2 (dataset transfer) due to resource overhead and feasibility concerns, while strongly supporting MI-Option 3 (model transfer) and MI-Option 4 (standardized reference models).
Open issues
- Network-controlled vs UE-driven model ID assignment approaches
- Prioritization and feasibility of different model identification options (MI-Options 1-4)
- Handling cross-vendor collaboration complexity in model transfer cases
- UE-side training data collection mechanisms and privacy considerations
Contributions
Discussion on AI/ML for beam management
China Telecom presents a comprehensive technical contribution on AI/ML for NR beam management, covering lifecycle management (LCM) aspects including data collection, model inference, and performance monitoring for both network-sided and UE-sided AI/ML models. The document contains 13 detailed proposals addressing…
China Telecom advocates for a comprehensive AI/ML beam management framework that supports both network-sided and UE-sided models with flexible data collection mechanisms. They push FOR beam prediction accuracy KPIs as the primary performance monitoring metric and support hybrid…
RAN2 inputs to TR 38.843
This is a liaison statement from RAN2 to RAN1 regarding UE-side data collection for UE-side model training in AI/ML for NR Air Interface work. The document contains no technical proposals but rather informs RAN1 that RAN2 has endorsed a Technical Paper (TP) on this topic and requests RAN1 to incorporate it into TR…
RAN2 is advocating for the inclusion of UE-side data collection capabilities for UE-side model training in AI/ML applications for the NR air interface, as evidenced by their endorsement of TP R2-2407807 and their request for RAN1 to incorporate this work into the technical…
AI/ML for Positioning Accuracy Enhancement
This Ericsson document presents a comprehensive analysis of AI/ML for positioning accuracy enhancement, providing 66 technical proposals across protocol integration, signaling enhancements, model outputs, training data collection, performance monitoring, and model inference. The document strongly advocates for…
Ericsson strongly advocates for sample-based measurements over legacy path-based measurements, demonstrating superior performance with lower complexity and signaling overhead. They firmly oppose including phase information as model input due to complexity, deployment costs, and…
Discussion on specification support for AI/ML-based beam management
This Futurewei document presents 8 proposals for specification support of AI/ML-based beam management in NR Release 19, covering performance monitoring, model inference, data collection, and assistance information aspects. The company advocates for reusing existing CSI frameworks as much as possible to minimize…
Futurewei strongly advocates for reusing existing CSI frameworks as much as possible to minimize specification effort and complexity. They oppose introducing new complex metrics like probability information and confidence levels (rejecting Opt 3 and Opt 4), preferring simpler,…
Discussion of additional study on AI/ML for NR air interface for CSI compression
Futurewei contributes to AI/ML for NR air interface CSI compression, discussing inter-vendor training collaboration options and providing temporal-domain CSI compression evaluation considering UCI loss. The document contains 10 proposals and 7 observations covering collaboration directions, quantization impacts,…
Futurewei advocates FOR using Rel-16 eType II codebook with new/enhanced parameters for data collection and monitoring to achieve better performance, even with additional overhead concerns. They push FOR comprehensive analysis of inter-vendor collaboration options through…
Discussion on other aspects of AI/ML model and data on AI/ML for NR air-interface
FUTUREWEI presents a comprehensive analysis of AI/ML model identification and data management for NR air interface, covering four model identification options, model transfer/delivery mechanisms, and inter-vendor collaboration approaches. The document contains 14 proposals and 3 observations addressing the…
FUTUREWEI advocates for simplified model identification approaches that minimize complexity while supporting two-sided models. They strongly push FOR network-controlled model ID assignment, cell-scoped associated IDs, and deprioritizing complex offline cross-vendor collaboration…
Discussion on AI/ML for beam management
This Huawei document presents a comprehensive analysis of AI/ML for beam management in NR, containing 38 detailed proposals and 9 observations covering data collection, inference procedures, performance monitoring, and UE capability reporting for both network-side and UE-side AI/ML models.
Huawei advocates FOR pragmatic reuse of existing CSI framework and signaling mechanisms while pushing AGAINST overly complex new procedures. They strongly support expanding beam measurement capabilities (up to 256 beams vs legacy 64) and unified solutions that work across…
Discussion on AI/ML for positioning accuracy enhancement
This Huawei document presents 25 proposals for AI/ML-based positioning accuracy enhancement in 5G NR, covering model input/output specifications, training procedures, consistency mechanisms, and lifecycle management across different positioning cases (Case 1: UE-based, Case 2: UE-assisted, Case 3: gNB-assisted).
Huawei advocates for pragmatic implementation-based solutions over rigid standardization, arguing that ambiguity issues can be resolved through consistent vendor implementations rather than tight specifications. They strongly oppose complex new signaling mechanisms, pushing…
Discussion on AI/ML for CSI prediction
Huawei argues against introducing associated IDs for ensuring consistency between training and inference in CSI prediction for AI/ML models, proposing instead UE-side performance monitoring approaches. The document contains 2 proposals and 5 observations addressing feasibility, necessity, and alternative solutions.
Huawei strongly advocates AGAINST introducing associated IDs or network-side indications for CSI prediction consistency, arguing they are unnecessary, technically infeasible, and create privacy risks. They push FOR UE-side performance monitoring as a sufficient…
Discussion on AI/ML for CSI compression
This Huawei document analyzes AI/ML for CSI compression in Release 19, focusing on inter-vendor training collaboration, temporal domain extensions, and remaining specification issues from Release 18. The document contains 17 proposals and 10 observations covering training methods, overhead concerns, monitoring…
Huawei advocates for Direction A (dataset/model sharing for UE-side offline engineering) over Direction B (direct parameter sharing) due to significantly lower air-interface overhead, supports prioritizing Option 4 (dataset sharing) and simpler sub-options (4-1, 4-2) over…
Discussion on other aspects of the additional study for AI/ML
This Huawei document discusses AI/ML air interface aspects focusing on model identification for two-sided models, model transfer/delivery, and UE-side training data collection. The document contains 8 proposals and 3 observations across multiple sections covering different model identification options and deployment…
Huawei advocates FOR focusing model identification discussions exclusively on two-sided models while eliminating MI-Option 1 from scope, and pushes FOR using Case y as baseline for UE-side training scenarios. They are AGAINST pursuing Case z1 for UE-side trained models due to…
LS on applicable functionality reporting for beam management UE-sided model
This is a liaison statement from RAN2 to RAN1 regarding AI/ML functionality reporting procedures for beam management UE-sided models, containing no specific proposals but presenting 10 detailed questions about the signaling procedure and functionality management.
RAN2 is seeking RAN1's technical input and validation on their agreed signaling procedures for AI/ML beam management functionality reporting, advocating for a structured 5-step process while acknowledging multiple technical aspects remain for further study (FFS). They are…
LS on AIML data collection
This liaison statement from RAN to SA groups presents 4 options for AI/ML data collection for UE-side model training, seeking SA input by RAN#106 to resolve consensus issues on MNO controllability and UP tunnel feasibility. The document contains no formal proposals but establishes requirements and analysis framework…
InterDigital is advocating for a comprehensive framework that balances MNO control with standardized AI/ML data collection, pushing for solutions that ensure full MNO controllability and visibility while maintaining privacy compliance. They are seeking SA group input to resolve…
Specification Support for AI/ML for Beam Management
This Kyocera document proposes comprehensive specification support for AI/ML-enabled beam management in 5G NR Rel-19, covering both network-sided and UE-sided models for spatial and temporal beam prediction. The document contains 24 proposals and 2 observations addressing configuration, reporting, consistency, and…
Kyocera advocates for a comprehensive AI/ML beam management framework that leverages existing CSI infrastructure while introducing minimal new specifications. They strongly support virtual Set A configuration for UE-side models, flexible Set B definition through new IEs, and…
Discussion on LS on applicable functionality reporting for beam management UE-sided model
Spreadtrum provides RAN1's response to RAN2's liaison statement on applicable functionality reporting for beam management UE-sided AI/ML models, presenting 8 comprehensive proposals addressing granularity, network-side conditions, configuration content, and activation mechanisms.
Spreadtrum advocates FOR mandatory network-side additional conditions and structured configuration frameworks, pushing for sub-use case level granularity and comprehensive inference configuration requirements. They are positioned AGAINST optional network-side conditions and…
Discussion on AIML for beam management
Spreadtrum presents their technical positions on AI/ML for beam management in NR, covering data collection, model inference, and performance monitoring aspects. The document contains 12 proposals and 3 observations addressing both UE-side and network-side AI/ML models for spatial and temporal beam prediction use cases.
Spreadtrum advocates for UE-initiated data collection for UE-side models, opposes larger quantization steps for inference to maintain accuracy, supports up to 16 beams reporting per instance, and favors reusing existing CSI frameworks to minimize specification impact. They push…
Discussion on AIML for CSI prediction
Spreadtrum proposes using associated IDs to ensure consistency between training and inference for CSI prediction in AI/ML-enhanced NR air interfaces. The document contains 2 main proposals focused on leveraging existing AI-BM conclusions and configuring associated IDs within the CSI framework.
Spreadtrum advocates FOR reusing existing AI-BM (beam management) conclusions and associated ID mechanisms for CSI prediction to minimize workload, and pushes AGAINST performance monitoring-based approaches (option 2) which would cause significant performance loss due to…
Discussion on AIML for CSI compression
Spreadtrum presents evaluation results for AI-based CSI Spatial-Temporal-Frequency (S-T-F) compression showing superior performance over Rel-16 eType II codebook, and provides 5 proposals addressing inter-vendor training collaboration, CQI determination, historical CSI handling, and performance monitoring for…
Spreadtrum strongly advocates FOR AI-based CSI Spatial-Temporal-Frequency (S-T-F) compression demonstrating it achieves over double the SGCS gain compared to spatial-frequency compression alone, and AGAINST option 5a for inter-vendor collaboration due to increased complexity.…
Discussion on other aspects of AI/ML model and data
Spreadtrum presents their views on AI/ML for NR air interface general aspects including data collection, model transfer/delivery, and model identification for two-sided models. The document contains 4 proposals and 4 observations addressing deprioritization of certain model transfer cases and procedures for model…
Spreadtrum advocates FOR deprioritizing model transfer cases z1 and z2 in Rel-19 due to cross-vendor collaboration burdens and proprietary design disclosure risks. They support mechanism 1a for UE data collection to avoid privacy exposure to network, and promote…
Specification support for beam management
This 3GPP RAN1 technical document from Tejas Networks presents 33 proposals and 6 observations for AI/ML-based beam management in 5G NR, covering both UE-sided and network-sided models for spatial (BM-Case1) and temporal (BM-Case2) beam prediction. The document addresses key aspects including Associated ID…
Tejas Networks advocates FOR simplified beam management through single resource set configurations where possible (Set B only for UE-sided BM-Case1), differential L1-RSRP reporting to reduce overhead, and flexible Associated ID mechanisms supporting both CSI framework and…
Specification support for positioning accuracy enhancement
This Tejas Networks document addresses AI/ML enhancements for NR positioning accuracy with 15 formal proposals and 14 observations covering model input parameters, training data collection, and performance monitoring across various positioning use cases.
Tejas Networks advocates FOR leveraging existing Release-17 path-based measurement frameworks and emphasizes the critical need to address receiver implementation dependencies in AI-ML model performance. They push AGAINST overly complex new measurement frameworks, instead…
Other aspects of AI/ML model and data
This 3GPP RAN1 document from Tejas Networks discusses AI/ML model identification and life cycle management for NR air interface, focusing on associated IDs for ensuring consistency between training and inference phases. The document contains 6 proposals and 4 observations covering model identification procedures,…
Tejas Networks advocates for UE-driven model ID assignment and supports maintaining current model transfer/delivery case prioritizations. They push FOR flexible many-to-one mapping between associated IDs and model IDs to enable generalized models, and FOR using existing RRC…
Draft reply LS on applicable functionality reporting for beam management UE-sided model
This is a reply liaison statement from vivo's RAN1 to RAN2 providing detailed answers to 10 questions about applicable functionality reporting for beam management UE-sided AI/ML models. The document contains technical clarifications but no formal proposals, focusing on the procedural aspects of AI/ML model deployment…
Vivo advocates FOR mandatory associated IDs as essential for maintaining consistency between AI/ML training and inference phases, pushing for FG-specific granularity in functionality definition. They are advocating AGAINST providing inference configuration in early steps,…
Discussion and reply LS on applicable functionality reporting for beam management UE-sided model
ZTE provides a comprehensive response to RAN2's liaison statement on AI/ML beam management functionality reporting, addressing 10 specific questions with 14 detailed proposals covering UE capability signaling, network-side conditions, and activation procedures.
ZTE advocates FOR flexible and optional network-side signaling to minimize overhead while maintaining functionality, supporting sub-use-case level granularity for AI/ML beam management, and reusing existing CSI framework mechanisms. They push AGAINST mandatory associated ID…
Discussion on AI/ML-based beam management
ZTE presents a comprehensive technical document on AI/ML-based beam management for 5G NR with approximately 35 proposals and 3 observations covering functionality-based lifecycle management, data collection enhancements, model inference mechanisms, performance monitoring, and network-side additional conditions…
ZTE strongly advocates FOR functionality-based lifecycle management over model-ID-based approaches, arguing it reduces complexity and leverages existing UE capability frameworks. They push FOR bitmap-based beam reporting methods to significantly reduce signaling overhead (up to…
Discussion on AI/ML-based positioning enhancement
ZTE presents a comprehensive technical document on AI/ML-based positioning enhancement for NR air interface with 30 proposals and 8 observations covering model training, inference, monitoring, and data collection aspects for different use cases.
ZTE advocates FOR maximizing reuse of existing 3GPP procedures and specifications rather than defining new enhancements, strongly supports CIR with phase information over PDP for better positioning accuracy despite higher overhead, pushes for sample-based measurements over…
Discussion on specification support for AI CSI prediction
ZTE proposes a methodical approach to AI CSI prediction consistency issues, advocating to first identify potential additional conditions before developing detailed solutions. The document contains 2 proposals focused on leveraging existing AI beam prediction mechanisms as a starting point.
ZTE advocates FOR a systematic approach that prioritizes identifying additional conditions before rushing into detailed consistency solutions, and FOR reusing proven AI beam prediction mechanisms. They are AGAINST introducing new associated IDs for scenarios/carrier frequency,…
Discussion on study for AI/ML CSI compression
ZTE provides comprehensive analysis on AI/ML CSI compression inter-vendor collaboration approaches, presenting 27 proposals across three main directions (UE-side offline engineering, on-device operation, fully standardized models) and addressing remaining specification issues including data collection, CQI…
ZTE advocates for prioritizing Option 3 (standardized reference model structure + parameter exchange) with NW-first training and over-the-air delivery as the most feasible inter-vendor collaboration approach. They push against Option 4 due to huge dataset exchange overhead and…
Discussion on other aspects of AI/ML model and data
ZTE's contribution analyzes AI/ML model identification and transfer options for NR air interface, comparing dataset transfer (MI-Option 2), model transfer (MI-Option 3), and standardized reference models (MI-Option 4). The document contains 12 proposals and 5 observations across model identification, model…
ZTE strongly advocates AGAINST MI-Option 2 (dataset transfer) due to feasibility concerns including huge resource overhead (1-10GB datasets vs 7-100MB models), large latency, and high UE power consumption. They strongly support MI-Option 3 (model transfer) and MI-Option 4…