| Challenge: | Existing safety alignment methods, such as RLHF, fall into a Safety-Utility Trade-off, resulting in severe over-rejection of benign household instructions. |
| Approach: | They propose a meta-cognitive Critical Agent that evaluates peer debates using a structured argumentation framework derived from the Toulmin Model. |
| Outcome: | The proposed architecture outperforms existing systems in the SafeAware-VH benchmark. |
Similar Papers
Debate, Deliberate, Decide (D3): A Cost-Aware Adversarial Framework for Reliable and Interpretable LLM Evaluation (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing evaluation tools for Large Language Models (LLMs) are inconsistency, bias, and lack of transparent decision criteria. |
| Approach: | They propose a cost-aware, adversarial multi-agent framework that orchestrates structured debate among role-specialized agents to produce reliable and interpretable evaluations. |
| Outcome: | The proposed framework orchestrates structured debate among role-specialized agents to produce reliable and interpretable evaluations. |
SELENE: Selective and Evidence-Weighted LLM Debating for Efficient and Reliable Reasoning (2026.eacl-industry)
Copied to clipboard
| Challenge: | Existing multi-agent debate frameworks are computationally expensive and prone to degradation under pro-longed debates due to redundant exchanges and unstable judging. |
| Approach: | They propose a framework that unifies Selective Debate Initiation (SDI) with Evidence Weighted Self-Consistency (EWSC) for adaptive, debate-on-demand reasoning. |
| Outcome: | Evaluated on BoolQ, CosmosQA, and an internal QnA benchmark, the proposed framework achieves higher factual robustness and efficiency. |
SafetyALFRED: Evaluating Safety-Conscious Planning of Vision Language Models (2026.findings-acl)
Copied to clipboard
Josue Torres-Fonseca, Naihao Deng, Yinpei Dai, Shane Storks, Yichi Zhang, Rada Mihalcea, Casey Kennington, Joyce Chai
| Challenge: | Existing safety evaluations focus on hazard recognition through disembodied question answering (QA) settings, but lack a critical gap in evaluating an agent. |
| Approach: | They evaluate multimodal large language models with six categories of kitchen hazards . they propose a safety-based approach that prioritizes multi-step corrective actions . |
| Outcome: | The proposed model can recognize hazards in QA settings, but average mitigation success rates are low . the proposed model is based on the embodied agent benchmark ALFRED . |
Free-MAD: Consensus-Free Multi-Agent Debate (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing multi-agent debate methods rely on multiple rounds of interaction among agents to reach consensus, and the final output is decided by majority voting in the last round. |
| Approach: | They propose a multi-agent debate framework that eliminates the need for consensus among agents and reconstructs the debate phase by introducing anti-conformity. |
| Outcome: | Experiments on eight benchmark datasets show that Free-MAD significantly improves reasoning performance while requiring only a single-round debate and thus reducing token costs. |
Beyond Reactive Safety: Risk-Aware LLM Alignment via Long-Horizon Simulation (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing alignment methods focus on reactive feedback, where immediate human perception is leveraged to judge sampled model responses as preference data for post-training. |
| Approach: | They propose a proof-of-concept framework that projects how model-generated advice could propagate through societal systems on a macroscopic scale over time, enabling more robust alignment. |
| Outcome: | The proposed framework achieves 20% improvement on existing safety benchmarks and an average win rate exceeding 70% against strong baselines. |
VestaBench: An Embodied Benchmark for Safe Long-Horizon Planning Under Multi-Constraint and Adversarial Settings (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Existing safety benchmarks do not represent a diverse range of multi-constraint tasks that require long-horizon planning with a focus on safety. |
| Approach: | They propose a benchmark to assess the safety of embodied AI agents under multiple constraints. |
| Outcome: | The proposed benchmarks show that LLMs perform poorly against their tasks . they also suffer significantly compromised safety outcomes . |
MALLM: Multi-Agent Large Language Models Framework (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Multi-agent debate (MAD) has demonstrated the ability to augment collective intelligence by scaling test-time compute and leveraging expertise. |
| Approach: | They propose an open-source framework that enables systematic analysis of multi-agent debates. |
| Outcome: | The proposed framework enables systematic analysis of multi-agent debate components. |
Debate to Align: Reliable Entity Alignment through Two-Stage Multi-Agent Debate (2026.findings-acl)
Copied to clipboard
| Challenge: | Entity alignment (EA) aims to identify entities referring to the same real-world object across different knowledge graphs (KGs). |
| Approach: | They propose a reliable EA framework based on multi-agent debate that improves embedding quality and introduces a two-stage multi-role debate mechanism to enhance reliability. |
| Outcome: | The proposed framework improves embedding quality and the reasoning capability of LLMs while enabling more efficient debate-based reasoning. |
M-MAD: Multidimensional Multi-Agent Debate for Advanced Machine Translation Evaluation (2025.acl-long)
Copied to clipboard
| Challenge: | Recent advances in large language models (LLMs) have shown their potential to deliver human-like judgments. |
| Approach: | They propose a systematic LLM-based multi-agent framework for advanced LLM as-a-judge MT evaluation that integrates dimension-specific results into a final evaluation judgment. |
| Outcome: | The proposed framework outperforms existing LLM-as-a-judge methods and competes with state-of-the-art automatic metrics even when powered by a suboptimal model like GPT-4o mini. |
SAGE: A Generic Framework for LLM Safety Evaluation (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Current safety evaluation methodologies focus on single-turn interactions with generic policies, failing to capture conversational dynamics of real-world usage and application-specific harms. |
| Approach: | They propose a framework for customized and dynamic harm evaluations that employs prompted adversarial agents with diverse personalities based on the Big Five model. |
| Outcome: | The proposed framework enables system-aware multi-turn conversations that adapt to target applications and harm policies. |