| Challenge: | Existing input-based routers optimize cost-performance trade-offs but provide no formal bound on how often cheaper model fails . a feasibility analysis across all 10 RouterBench models reveals that routability is jointly model- and task-dependent. |
| Approach: | They adapt a proactive conformal gate framework to LLM routing to provide cost-performance trade-offs . they find that a third axis is missing: the routability of queries is model-dependent . |
| Outcome: | The proposed method maintains the target within the tolerance across two benchmarks . it is the first input-based LLM router with distribution-free safety guarantees . |
Similar Papers
How Robust Are Router-LLMs? Analysis of the Fragility of LLM Routing Capabilities (2026.eacl-long)
Copied to clipboard
| Challenge: | Large language model (LLM) routing has emerged as a promising solution to balancing computational costs and performance. |
| Approach: | They propose a framework that categorizes router performance across a broad spectrum of query types . large language models have revolutionized natural language processing . |
| Outcome: | The proposed framework categorizes router performance across a broad spectrum of query types . it integrates privacy and safety assessments to reveal hidden risks . |
InferenceDynamics: Adaptive LLM Routing through Structured Capability and Knowledge Profiling (2026.acl-long)
Copied to clipboard
Haochen Shi, Tianshi Zheng, Weiqi Wang, Baixuan Xu, Chunyang Li, Chunkit Chan, Tao Fan, Yangqiu Song
| Challenge: | Large Language Model (LLM) routing is a pivotal technique for navigating a diverse landscape of LLMs. |
| Approach: | They propose a flexible and scalable multi-dimensional routing framework that models the capability and knowledge of models. |
| Outcome: | The proposed framework can be used to generalize and identify top-performing models for group-level routing using modern benchmarks including MMLU-Pro, GPQA, BigGenBench, and LiveBench. |
X-Router: Decoupling Knowledge and Reasoning for Cost-Effective LLM Inference (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing adaptive methods focus on a single axis, overlooking evidence need and reasoning depth are only partially correlated. |
| Approach: | They propose a dual-axis routing framework that separates retrieval necessity from reasoning necessity under a user-defined cost–quality trade-off. |
| Outcome: | The proposed framework reduces token usage and latency while improving answer quality over strong baselines. |
MixLLM: Dynamic Routing in Mixed Large Language Models (2025.naacl-long)
Copied to clipboard
Xinyuan Wang, Yanchi Liu, Wei Cheng, Xujiang Zhao, Zhengzhang Chen, Wenchao Yu, Yanjie Fu, Haifeng Chen
| Challenge: | Large Language Models (LLMs) exhibit potential artificial generic intelligence, however, their usage is costly with high response latency. |
| Approach: | They develop a dynamic contextual-bandit-based routing system for query-LLM assignment that leverages query tags to enhance query embeddings. |
| Outcome: | The proposed model maximizes response quality and minimizes cost and latency. |
LLMRouterBench: A Massive Benchmark and Unified Framework for LLM Routing (2026.findings-acl)
Copied to clipboard
Hao Li, Yiqun Zhang, Zhaoyan Guo, Chenxu Wang, Shengji Tang, Qiaosheng Zhang, Yang Chen, Biqing Qi, Peng Ye, Lei Bai, Zhen Wang, Shuyue Hu
| Challenge: | Large language model (LLM) routing assigns each query to the best suitable model from an ensemble. |
| Approach: | They introduce a large-scale benchmark and unified framework for LLM routing . they find that many routing methods exhibit similar performance under unified evaluation . |
| Outcome: | The proposed benchmark provides comprehensive metrics for both performance-oriented and performance-cost trade-off routing. |
Into the Gray Zone: Domain Contexts Can Blur LLM Safety Boundaries (2026.acl-long)
Copied to clipboard
Ki Sen Hung, Xi Yang, Chang Liu, Haoran Li, Kejiang Chen, Changxuan Fan, Tsun On Kwok, Weiming Zhang, Xiaomeng Li, Yangqiu Song
| Challenge: | a goal of LLM alignment is to balance usefulness with harmlessness, but this conflictes when knowledge serves both legitimate and malicious purposes. |
| Approach: | They propose a framework that combines safety-research contexts with adversarial interactions to exploit a vulnerability in Jargon queries. |
| Outcome: | a framework outperforms existing methods in analyzing Jargon queries, a study shows . it achieves 93% of attacks across seven models, while remaining useful, the authors say . |
LLM-VA: Resolving the Jailbreak-Overrefusal Trade-off via Vector Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Existing vector steering methods adjust the magnitude of answer vectors, but this creates a fundamental trade-off—reducing jailbreak increases over-refusal. |
| Approach: | They propose a method which aligns va with vb through closed-form weight updates, making the model’s willingness to respond causally dependent on its safety assessment. |
| Outcome: | Experiments on 12 LLMs show that the proposed method achieves 11.45% higher F1 than the best baseline while preserving 95.92% utility. |
Safety Is Not Universal: The Selective Safety Trap in LLM Alignment (2026.findings-acl)
Copied to clipboard
Iago Alves Brito, Walcy Rios, Julia Soares Dollis, Diogo Fernandes Costa Silva, Arlindo Rodrigues Galvão Filho
| Challenge: | Existing safety evaluations of large language models aggregate harms under generic categories such as "Identity Hate" a bilingual benchmark identifies a selective safety trap, where defense rates vary by up to 42% within the same model solely based on the target group. |
| Approach: | They propose a bilingual adversarial benchmark to audit selective safety in large language models . defense rates vary by up to 42% within the same model solely based on target group . |
| Outcome: | The proposed benchmark identifies a selective safety trap in large language models . defense rates vary by up to 42% within the same model solely based on the target group. |
SafeConstellations: Mitigating Over-Refusals in LLMs Through Task-Aware Representation Steering (2026.acl-long)
Copied to clipboard
| Challenge: | Current safety alignment methods fail to identify intended benign task before refusing to respond. |
| Approach: | They propose a method that uses inference-time trajectory-shifting to guide model behavior . they show that LLMs persist in refusing inputs containing harmful content . |
| Outcome: | The proposed approach reduces over-refusals with minimal impact on utility. |
Course-Correction: Safety Alignment Using Synthetic Preferences (2024.emnlp-industry)
Copied to clipboard
Rongwu Xu, Yishuo Cai, Zhenhong Zhou, Renjie Gu, Haiqin Weng, Liu Yan, Tianwei Zhang, Wei Xu, Han Qiu
| Challenge: | Recent studies show that large language models generate harmful content, but the potential for generating harmful content is an escalating concern. |
| Approach: | They propose to fine-tune LLMs with preference learning to emphasize the preference for timely course-correction by using an automated pipeline. |
| Outcome: | The proposed model improves course-correction skills without affecting general performance and resists jailbreak attacks. |