# UTS at PsyDefDetect: Multi-Agent Councils and Absence-Based Reasoning for Defense Mechanism Classification

arXiv: https://arxiv.org/abs/2605.09769 · HTML: https://mlsanity.com/paper/2605.09769/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 6 of its 21 references that are in the map.
- First posted on arXiv 10 May 2026.

## Builds on (most cited first, 6 of 6)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Improving Factuality and Reasoning in Language Models through Multiagent Debate](https://mlsanity.com/paper/2305.14325/) (2023)
- [Energy and Policy Considerations for Deep Learning in NLP](https://mlsanity.com/paper/1906.02243/) (2019)
- [Text Embeddings by Weakly-Supervised Contrastive Pre-training](https://mlsanity.com/paper/2212.03533/) (2022)
- [AI safety via debate](https://mlsanity.com/paper/1805.00899/) (2018)
- [You Never Know a Person, You Only Know Their Defenses: Detecting Levels of Psychological Defense Mechanisms in Supportive Conversations](https://mlsanity.com/paper/2512.15601/) (2025)

## Built on by (most cited first, 1 of 1)

- [Overview of the PsyDefDetect Shared Task at BioNLP 2026: Detecting Levels of Psychological Defense Mechanisms in Supportive Conversations](https://mlsanity.com/paper/2605.24907/) (2026)

## Abstract (arXiv)

This paper describes our system for classifying psychological defense mechanisms in emotional support dialogues using the Defense Mechanism Rating Scales (DMRS), placing second (F1 0.406) among 64 teams. A central insight is that defense mechanisms are defined by what is absent: missing affect, blocked cognition, denied reality. We encode this as an affect-cognition integration spectrum in prompt-level clinical rules, which account for the largest single gain (+11.4pp F1). Our architecture is a multi-phase deliberative council of Gemini 2.5 agents where class-specific advocates rate evidence strength rather than voting, achieving F1 0.382 with no fine-tuning - a top-5 result on its own. We find, however, that the council is confidently wrong about minority classes: 59-80% of stable minority predictions are incorrect, driven by a systematic "L7 attractor" in which emotional content defaults to the majority class. A targeted override ensemble from three fine-tuned Qwen3.5 models applies 16 overrides (+2.4pp), selected by a structured multi-agent system (builder, critic, regression guard) that produced a larger F1 gain in one iteration than 8 prior attempts combined.