# Local Gains and Fixed-Assignment Set Losses in Shared Set Decoders

arXiv: https://arxiv.org/abs/2608.14717 · HTML: https://mlsanity.com/paper/2608.14717/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 19 references that are in the map.
- First posted on arXiv 12 Aug 2026.

## Builds on (most cited first, 14 of 14)

- [End-to-End Object Detection with Transformers](https://mlsanity.com/paper/2005.12872/) (2020)
- [DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection](https://mlsanity.com/paper/2203.03605/) (2022)
- [Attention is not Explanation](https://mlsanity.com/paper/1902.10186/) (2019)
- [Attention is not not Explanation](https://mlsanity.com/paper/1908.04626/) (2019)
- [DN-DETR: Accelerate DETR Training by Introducing Query DeNoising](https://mlsanity.com/paper/2203.01305/) (2022)
- [Towards Best Practices of Activation Patching in Language Models: Metrics and Methods](https://mlsanity.com/paper/2309.16042/) (2023)
- [Group DETR: Fast DETR Training with Group-Wise One-to-Many Assignment](https://mlsanity.com/paper/2207.13085/) (2022)
- [Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching](https://mlsanity.com/paper/2311.17030/) (2023)
- [Detection Transformer with Stable Matching](https://mlsanity.com/paper/2304.04742/) (2023)
- [Relation DETR: Exploring Explicit Position Relation Prior for Object Detection](https://mlsanity.com/paper/2407.11699/) (2024)
- [Causal Abstraction with Soft Interventions](https://mlsanity.com/paper/2211.12270/) (2022)
- [Causality is Key for Interpretability Claims to Generalise](https://mlsanity.com/paper/2602.16698/) (2026)
- [Dual-R-DETR: Resolving Query Competition with Pairwise Routing in Transformer Decoders](https://mlsanity.com/paper/2512.13876/) (2025)
- [Beyond Importance: Interchange-Sobol Sensitivity Reveals Task-Specific Content Channels in Transformer Components](https://mlsanity.com/paper/2606.20678/) (2026)

## Abstract (arXiv)

A query-relation deletion can improve the edited slot while reducing the utility of the prediction set that contains it. We study this tension in two related ResNet-50 DETR-family checkpoints using recorded, selection-conditional evidence from 710 paired image-relation units per checkpoint. The primary comparison subtracts a matched active control, which deletes the same leader source at a different recorded recipient, from the selected target deletion. It is therefore a composite contrast rather than a same-recipient placebo. The target-minus-control contrast is locally positive and fixed-assignment negative in both checkpoints. The opposite-sign pattern occurs within 302/710 DETR units and 460/710 DINO units. After rematching, the corresponding counts are 285/710 and 433/710. Rematching and native selection absorb enough of the mean loss for DETR intervals to cross zero, whereas DINO intervals remain negative, so persistence across readouts differs by checkpoint. A fixed-map comparison between hard deletion and a mass-preserving edit also differs before rematching. That comparison is conditional on the outcome-blind map and does not establish same-dose transport. Local intervention success therefore does not determine the consequence for a jointly decoded set. The supported conclusion is selection-conditional deletion sensitivity whose persistence depends on the readout and intervention operator. We do not identify an intervention-invariant edge mechanism, detector-level degradation, population prevalence, or the value of a training-time regularizer.