Position changes

CATT strengthened

CATT refined their stance on associated ID configuration, specifying it should be at the CSI-Report level rather than requiring similar DL Tx beam properties. They hardened their opposition to extending Rel-17 TCI state activation for multiple future time instances in BM-Case 2, citing complexity. New technical arguments were added regarding the benefit of aligning Rx beam information for NW-sided models and the introduction of an enhanced CPU pool distinct from legacy CPUs. Their support for event-based performance monitoring linked to inference reports is a new addition.

FUTUREWEI strengthened

FUTUREWEI hardened their opposition to supporting Opt 3 and Opt 4 for UE-sided model inference reports, explicitly arguing that defining probability and confidence metrics is difficult and offers unclear benefits. They added a new requirement to increase the maximum number of reported beams (M) for network-sided models from 4 to 8. Their support for using RS ID as an implicit beam ID indicator is preserved. They added a preference for Option B for RSRP reporting, which prioritizes measured L1-RSRP over predicted values when measurements are available.

Google shifted

Google shifted their focus from general flexible channel measurements to specifically proposing the extension of enhanced path-based measurement to the UE side to reduce reporting overhead. They added a new proposal to report L1-SINR alongside path-based measurements to enable network filtering. For model performance monitoring in Case 1, they refined their support for UE-side monitoring (Option A) by specifying a simple 1-bit failure indication and explicitly opposing Option B due to functional redundancy and privacy concerns. They also added support for Alternative 4, requiring explicit provision of TRP geographical coordinates from the LMF.

Huawei shifted

Huawei shifted their argument against phase information, now citing that double phase difference cannot mitigate phase errors in NLOS scenarios and that timing/power suffices. They added a new technical case against the associated ID for TRP location consistency, arguing that TRP locations change infrequently and UE-side burden from combinatorial model training would be excessive. They preserved their stance on reusing legacy IEs for LOS/NLOS indicators and added a proposal to distinguish Rel-19 timing information via timing quality indicators or a specific Rel-19 type indicator.

Ericsson strengthened

Ericsson hardened their stance on measurement inputs, moving from a compromise position supporting both sample-based and path-based measurements to explicitly opposing multi-port PDP due to doubled signaling size and marginal performance gains. They consolidated their technical case against phase information and CIR inputs, citing random initial phase issues and high signaling overhead. Additionally, they refined the monitoring framework by establishing self-monitoring as the baseline, with the inference entity responsible for metric calculation, rather than the broader label-free self-monitoring proposal from the prior meeting.

CATT shifted

CATT preserved its stance on negligible impact of tilt/TXRU but refined its position by adding specific proposals for a new processing unit type or enhanced CPU separate from the legacy pool. They added a requirement to distinguish AI/ML CSI reports via a new report quantity or identifier. Their monitoring preference was consolidated to explicitly prefer SGCS for Type 1 and Type 3 while deprioritizing Type 2, moving from general monitoring advocacy to specific metric and resource definitions.

CMCC shifted

CMCC softened its stance on associated IDs, moving from proposing a combined solution to preferring no specification of consistency unless degradation is proven. They refined their Type 3 monitoring proposal by explicitly naming SGCS and NMSE as intermediate KPIs. They added new specifics for Type 1 monitoring, proposing two alternatives: calculation/comparison outcomes based on thresholds or recommended LCM decisions. The position shifted from exploring ID mechanisms to focusing on conditional consistency and specific monitoring outputs.

Google shifted

Google shifted from proposing specific associated ID management to supporting the reuse of the broader Beam Management inference and monitoring frameworks. They added a new specific metric proposal: CQI offset between predicted and ground-truth CSI. They refined their resource configuration proposal by specifying separate CSI-RS resource sets for inference and monitoring. They also added a preference for NW-triggered reports via UCI while deprioritizing event-triggered reports.

Huawei shifted

Huawei preserved its core argument against network-side associated IDs but refined its monitoring proposal by adding specific priority rules where monitoring reports take precedence over inference reports. They added new technical requirements for separating CPU counting between legacy and AI/ML CSI reporting and advocating for L1 signaling for monitoring results to address latency. The position shifted from general feasibility arguments to specific signaling layer (L1) and resource management (CPU separation) proposals.

LG Electronics shifted

LG Electronics refined its position from high-level consistency arguments to specific implementation mechanisms, adding proposals for reusing the typeIIDoppler-r18' codebook and the existing CPU mechanism. They preserved their opposition to Type 2 monitoring overhead but added a new technical constraint regarding the specification of time limits or buffering windows for inference results linked to monitoring reports. The earlier focus on tilt angle consensus was dropped in favor of these concrete resource and timing definitions.

Ericsson strengthened

Ericsson hardened its opposition to consistency enhancements by consolidating per-aspect arguments into a blanket conclusion of inapplicability. They refined their proposal by specifying the reuse of the functionality-based LCM framework and requiring the typeII-Doppler-r18' codebook format for both predicted and ground-truth CSI. They narrowed the intermediate KPI scope to SGCS only, explicitly excluding NMSE, and maintained their opposition to Type 2 monitoring.

Spreadtrum shifted

Spreadtrum shifted from emphasizing UE-initiated control to favoring network-provided configurations while maintaining opposition to complex new metrics.

Huawei strengthened

Huawei shifted from a conservative approach emphasizing reuse of existing frameworks to a more aggressive stance pushing for significant enhancements and expanded capabilities.

Huawei strengthened

Huawei strengthened their position by expanding opposition from general complex signaling to specifically opposing tight specification of sample-based measurements and mandatory phase information.

ZTE strengthened

ZTE strengthened their position by providing specific quantitative evidence (1.2-2.2x performance improvement) to support their advocacy for CIR with phase information.

ZTE strengthened

ZTE evolved from a systematic approach with conditional support to a stronger opposition stance, now emphasizing model generalization capabilities as sufficient without additional consistency mechanisms.

Huawei strengthened

Huawei maintained their support for Direction A but strengthened their position by specifically favoring Option 4-1 and adding strong opposition to Direction C based on practical implementation concerns.

Huawei strengthened

Huawei expanded their opposition to cross-vendor collaboration from just Case z1 to Cases z1, z2, z3, and z5, taking a stronger stance against complexity.

ZTE strengthened

ZTE provided more specific technical justification for opposing MI-Option 2 with concrete resource overhead figures (1-10GB transfers) and added explicit support for Case z4.