A per-token feature from temperature-induced changes in LLM token distributions predicts within-prompt creativity rank at Spearman rho 0.918 vs LLM judges and 0.870 vs humans, outperforming perplexity, entropy, top-1 margin, and compression baselines.
hub
A Diversity-Promoting Objective Function for Neural Conversation Models
14 Pith papers cite this work. Polarity classification is still indexing.
14
Pith papers citing it
abstract
Sequence-to-sequence neural network models for generation of conversational responses tend to generate safe, commonplace responses (e.g., "I don't know") regardless of the input. We suggest that the traditional objective function, i.e., the likelihood of output (response) given input (message) is unsuited to response generation tasks. Instead we propose using Maximum Mutual Information (MMI) as the objective function in neural models. Experimental results demonstrate that the proposed MMI models produce more diverse, interesting, and appropriate responses, yielding substantive gains in BLEU scores on two conversational datasets and in human evaluations.
hub tools
citation-role summary
background 3
dataset 1
citation-polarity summary
representative citing papers
Autoregressive transformers exhibit measurable cognitive fatigue during extended generation, quantified by the Fatigue Index that predicts degradation (AUROC 0.95) and repetition (rho 0.94).
Introduces TBPO, which derives a Bregman-divergence density-ratio matching objective for token-level preference optimization that generalizes DPO while preserving the induced optimal policy.
REFORM uses reward-guided controlled decoding to generate preference-class-consistent responses that the reward model mis-scores, then retrains the reward model on these failure modes to improve robustness.
The authors generate and publicly release the first large-scale open dataset of three million structured moral fables produced by small open language models together with a reproducible LLM-judge evaluation pipeline.
Distinct Leaf Enumeration (DLE) replaces stochastic self-consistency sampling with deterministic traversal of a truncated decoding tree to enumerate distinct leaves, increasing coverage and reducing redundant computation while improving performance on math, coding, and reasoning benchmarks.
A DQN-style planner initialized from a pretrained LLM selects the system's emotion before response generation, using a hybrid reward of dataset imitation and GPT-4o-scored Plutchik theory, enabling streaming emotional TTS.
Nano-EmoX is a compact 2.2B multimodal model that unifies six core affective tasks across perception, understanding, and interaction levels via a curriculum framework, achieving competitive benchmark performance.
Summing outputs from separately trained QLoRA PEFT modules provides strong performance for attribute-controlled text generation, often matching or exceeding single-task modules even on single-attribute tests.
LaMDA shows that fine-tuning on human-value annotations and consulting external knowledge sources significantly improves safety and factual grounding in large dialog models beyond what scaling alone achieves.
A hypernetwork produces a condition-dependent beta that meta-gates SwiGLU nonlinearity, giving LLMs adaptive behavior across task, domain, persona and style inputs without finetuning.
RAFT aligns generative models by ranking samples with a reward model and fine-tuning only on the top-ranked outputs, reporting gains on reward scores and automated metrics for LLMs and diffusion models.
Re-evaluating controlled text generation systems under standardized conditions reveals that many published performance claims do not hold, highlighting the need for consistent evaluation practices.
Mainstream conversational models show escalating affective misalignments and ethical guidance failures during staged emotional trajectories, organized into a taxonomy of interactional breakdowns.
citing papers explorer
-
Before and After Temperature: A Distributional View of Creative LLM Generation
A per-token feature from temperature-induced changes in LLM token distributions predicts within-prompt creativity rank at Spearman rho 0.918 vs LLM judges and 0.870 vs humans, outperforming perplexity, entropy, top-1 margin, and compression baselines.
-
Cognitive Fatigue in Autoregressive Transformers: Formalization and Measurement
Autoregressive transformers exhibit measurable cognitive fatigue during extended generation, quantified by the Fatigue Index that predicts degradation (AUROC 0.95) and repetition (rho 0.94).
-
TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching
Introduces TBPO, which derives a Bregman-divergence density-ratio matching objective for token-level preference optimization that generalizes DPO while preserving the induced optimal policy.
-
Teach a Reward Model to Correct Itself: Reward Guided Adversarial Failure Discovery for Robust Reward Modeling
REFORM uses reward-guided controlled decoding to generate preference-class-consistent responses that the reward model mis-scores, then retrains the reward model on these failure modes to improve robustness.
-
TF1-EN-3M: Three Million Synthetic Moral Fables for Training Small, Open Language Models
The authors generate and publicly release the first large-scale open dataset of three million structured moral fables produced by small open language models together with a reproducible LLM-judge evaluation pipeline.
-
Efficient Test-Time Inference via Deterministic Exploration of Truncated Decoding Trees
Distinct Leaf Enumeration (DLE) replaces stochastic self-consistency sampling with deterministic traversal of a truncated decoding tree to enumerate distinct leaves, increasing coverage and reducing redundant computation while improving performance on math, coding, and reasoning benchmarks.
-
Self-EmoQ: Plutchik-Guided Value-based Planning to Drive Streaming Emotional TTS
A DQN-style planner initialized from a pretrained LLM selects the system's emotion before response generation, using a hybrid reward of dataset imitation and GPT-4o-scored Plutchik theory, enabling streaming emotional TTS.
-
Nano-EmoX: Unifying Multimodal Emotional Intelligence from Perception to Empathy
Nano-EmoX is a compact 2.2B multimodal model that unifies six core affective tasks across perception, understanding, and interaction levels via a curriculum framework, achieving competitive benchmark performance.
-
Output Composability of QLoRA PEFT Modules for Plug-and-Play Attribute-Controlled Text Generation
Summing outputs from separately trained QLoRA PEFT modules provides strong performance for attribute-controlled text generation, often matching or exceeding single-task modules even on single-attribute tests.
-
LaMDA: Language Models for Dialog Applications
LaMDA shows that fine-tuning on human-value annotations and consulting external knowledge sources significantly improves safety and factual grounding in large dialog models beyond what scaling alone achieves.
-
Learn-To-Learn on Arbitrary Textual Conditioning: A Hypernetwork-Driven Meta-Gated LLM
A hypernetwork produces a condition-dependent beta that meta-gates SwiGLU nonlinearity, giving LLMs adaptive behavior across task, domain, persona and style inputs without finetuning.
-
RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment
RAFT aligns generative models by ranking samples with a reward model and fine-tuning only on the top-ranked outputs, reporting gains on reward scores and automated metrics for LLMs and diffusion models.
-
A Comparative Study of Controlled Text Generation Systems Using Level-Playing-Field Evaluation Principles
Re-evaluating controlled text generation systems under standardized conditions reveals that many published performance claims do not hold, highlighting the need for consistent evaluation practices.
-
Breakdowns in Conversational AI: Interactional Failures in Emotionally and Ethically Sensitive Contexts
Mainstream conversational models show escalating affective misalignments and ethical guidance failures during staged emotional trajectories, organized into a taxonomy of interactional breakdowns.