ConceptMath: A Bilingual Concept-wise Benchmark for Measuring Mathematical Reasoning of Large Language Models

This paper introduces ConceptMath, a bilingual (English and Chinese), fine-grained benchmark that evaluates concept-wise mathematical reasoning of Large Language Models (LLMs). Unlike traditional benchmarks that evaluate general mathematical reasoning with an average accuracy, ConceptMath systematically organizes math problems under a hierarchy of math concepts, so that mathematical reasoning can be evaluated at different granularity with concept-wise accuracies. Based on our ConcepthMath, we evaluate a broad range of LLMs, and we observe existing LLMs, though achieving high average accuracies on traditional benchmarks, exhibit significant performance variations across different math concepts and may even fail catastrophically on the most basic ones. Besides, we also introduce an efficient fine-tuning strategy to enhance the weaknesses of existing LLMs. Finally, we hope ConceptMath could guide the developers to understand the fine-grained mathematical abilities of their models and facilitate the growth of foundation models.

MT-Bench-101: A Fine-Grained Benchmark for Evaluating Large Language Models in Multi-Turn Dialogues

The advent of Large Language Models (LLMs) has drastically enhanced dialogue systems. However, comprehensively evaluating the dialogue abilities of LLMs remains a challenge. Previous benchmarks have primarily focused on single-turn dialogues or provided coarse-grained and incomplete assessments of multi-turn dialogues, overlooking the complexity and fine-grained nuances of real-life dialogues. To address this issue, we introduce MT-Bench-101, specifically designed to evaluate the fine-grained abilities of LLMs in multi-turn dialogues. By conducting a detailed analysis of real multi-turn dialogue data, we construct a three-tier hierarchical ability taxonomy comprising 4208 turns across 1388 multi-turn dialogues in 13 distinct tasks. We then evaluate 21 popular LLMs based on MT-Bench-101, conducting comprehensive analyses from both ability and task perspectives and observing differing trends in LLMs performance across dialogue turns within various tasks. Further analysis indicates that neither utilizing common alignment techniques nor chat-specific designs has led to obvious enhancements in the multi-turn abilities of LLMs. Extensive case studies suggest that our designed tasks accurately assess the corresponding multi-turn abilities.

Emulated Disalignment: Safety Alignment for Large Language Models May Backfire!

Large language models (LLMs) need to undergo safety alignment to ensure safe conversations with humans. However, in this work, we introduce an inference-time attack framework, demonstrating that safety alignment can also unintentionally facilitate harmful outcomes under adversarial manipulation. This framework, named Emulated Disalignment (ED), adversely combines a pair of open-source pre-trained and safety-aligned language models in the output space to produce a harmful language model without additional training. Our experiments with ED across three datasets and four model families (Llama-1, Llama-2, Mistral, and Alpaca) show that ED doubles the harmfulness of pre-trained models and outperforms strong baselines, achieving the highest harmful rate in 43 out of 48 evaluation subsets by a large margin. Crucially, our findings highlight the importance of reevaluating the practice of open-sourcing language models even after safety alignment.

Attacks, Defenses and Evaluations for LLM Conversation Safety: A Survey

Large Language Models (LLMs) are now commonplace in conversation applications. However, their risks of misuse for generating harmful responses have raised serious societal concerns and spurred recent research on LLM conversation safety. Therefore, in this survey, we provide a comprehensive overview of recent studies, covering three critical aspects of LLM conversation safety: attacks, defenses, and evaluations. Our goal is to provide a structured summary that enhances understanding of LLM conversation safety and encourages further investigation into this important subject. For easy reference, we have categorized all the studies mentioned in this survey according to our taxonomy, available at: https://github.com/niconi19/LLM-conversation-safety.

Beyond One-Preference-for-All: Multi-Objective Direct Preference Optimization for Language Models

A single language model (LM), despite aligning well with an average labeler through reinforcement learning from human feedback (RLHF), may not universally suit diverse human preferences. Recent approaches thus pursue customization, training separate principle-based reward models to represent different alignment objectives (e.g. helpfulness, harmlessness, or honesty). Different LMs can then be trained for different preferences through multi-objective RLHF (MORLHF) with different objective weightings. Yet, RLHF is unstable and resource-heavy, especially for MORLHF with diverse and usually conflicting objectives. In this paper, we present Multi-Objective Direct Preference Optimization (MODPO), an RL-free algorithm that extends Direct Preference Optimization (DPO) for multiple alignment objectives. Essentially, MODPO folds LM learning directly into reward modeling, aligning LMs with the weighted sum of all principle-based rewards using pure cross-entropy loss. While theoretically guaranteed to produce the same optimal solutions as MORLHF, MODPO is practically more stable and computationally efficient, obviating value function modeling and online sample collection. Empirical results in safety alignment and long-form question answering confirm that MODPO matches or outperforms existing methods, consistently producing one of the most competitive LM fronts that cater to diverse preferences with 3 times fewer computations compared with MORLHF.

Beyond One-Preference-for-All: Multi-Objective Direct Preference Optimization

Language models (LMs), despite aligning well with an average labeler through reinforcement learning from human feedback (RLHF), may not universally suit diverse human preferences. Recent approaches therefore opt for customization by collecting multi-dimensional feedback and creating distinct rewards for each dimension (e.g., helpfulness, harmlessness, honesty). LMs can then be tailored to different preferences using multi-objective RL (MORL) with different reward weightings. Yet, RL fine-tuning is unstable and resource-heavy, especially for MORLHF with diverse and usually conflicting objectives. In this paper, we present Multi-Objective Direct Preference Optimization (MODPO), an RL-free algorithm that extends Direct Preference Optimization (DPO) for multiple alignment objectives. Essentially, MODPO trains different LMs to represent different collective reward models that combine all objectives with specific weightings. With a simple cross-entropy loss, the LMs optimized against the MODPO objective are analytically the exact solutions of the original MORLHF objective. Empirical results in safety alignment and long-form question answering confirm that MODPO matches or outperforms existing methods, efficiently producing a Pareto-optimal set of LMs that cater to diverse preferences with 3 times less computational resources compared with MORLHF.