相关论文: Where does output diversity collapse in post-train…

Instruction-tuning large language models (LLMs) reduces the diversity of their outputs, which has implications for many tasks, particularly for creative tasks. This paper investigates the ``diversity gap'' for a writing prompt narrative…

Instruction-tuned large language models (LLMs) employ structured templates, such as role markers and special tokens, to enforce format consistency during inference. However, we identify a critical limitation of such formatting: it induces a…

As creative writing tasks do not have singular correct answers, large language models (LLMs) trained to perform these tasks should be able to generate diverse valid outputs. However, LLM post-training often focuses on improving generation…

Recent advances in post-training techniques have endowed Large Language Models (LLMs) with enhanced capabilities for tackling complex, logic-intensive tasks through the generation of supplementary planning tokens. This development raises a…

RL-trained Lean theorem provers mode-collapse at inference time: on miniF2F-test with DeepSeek-Prover-V1.5-RL, doubling the i.i.d.\ sampling budget from $k{=}32$ to $k{=}64$ produces zero additional solved theorems (42/244 in both cases). A…

Output diversity is crucial for Large Language Models as it underpins pluralism and creativity. In this work, we reveal that controlling the language used during model thinking-the language of thought-provides a novel and structural source…

Large language models produce fluent fiction, yet their creative output is widely seen as flat. We ask where this quality originates in the training and whether it affects different domains of human fiction equally. We construct a matched…

In various fields of knowledge creation, including science, new ideas often build on pre-existing information. In this work, we explore this concept within the context of language models. Specifically, we explore the potential of…

On-policy reinforcement learning methods like GRPO suffer from mode collapse: they exhibit reduced solution diversity, concentrating probability mass on a single solution once discovered and ceasing exploration of alternative strategies. We…

Reinforcement learning (RL) has emerged as a popular method for post-training large language models (LLMs). While improving the model's performance on downstream tasks, it often reduces the model's output diversity, leading to narrow,…

In modern LLMs, linguistic features function not as stylistic artifacts but as probes of probability mass, allocated under training alignment objectives. Language models trained with contemporary pipelines exhibit severe reshaping of…

Post-training (via supervised fine-tuning) improves instruction-following, but often induces semantic mode collapse by biasing models toward low-entropy fine-tuning data at the expense of the high-entropy pretraining distribution.…

Diversity is essential for language-model applications ranging from creative generation to scientific discovery, yet modern LLMs often collapse into a narrow subset of plausible outputs. While prior work has developed benchmarks for…

The remarkable capabilities of modern large reasoning models are largely unlocked through post-training techniques such as supervised fine-tuning (SFT) and reinforcement learning (RL). However, the architectural mechanisms behind such…

Reasoning-enhanced large language models (RLLMs), whether explicitly trained for reasoning or prompted via chain-of-thought (CoT), have achieved state-of-the-art performance on many complex reasoning tasks. However, we uncover a surprising…

Reinforcement learning (RL) has emerged as a powerful method for improving the reasoning abilities of large language models (LLMs). Outcome-based RL, which rewards policies solely for the correctness of the final answer, yields substantial…

Supervised and preference-based fine-tuning techniques have become popular for aligning large language models (LLMs) with user intent and correctness criteria. However, real-world training data often exhibits spurious correlations --…

When instructed to underperform on multiple-choice evaluations, do language models engage with question content or fall back on positional shortcuts? We map the boundary between these regimes using a six-condition adversarial…

Reinforcement learning with verifiable reward (RLVR) has been instrumental in eliciting strong reasoning capabilities from large language models (LLMs) via long chains of thought (CoT). During RLVR training, we formalize and systemically…

Reasoning large language models achieve impressive test-time scaling by thinking for longer, but this performance gain comes at significant compute cost. Directly limiting test-time budget hurts overall performance, but not all problems are…