Line of inquiry
Inquiring lines›How can we understand and verify l…›How do language models construct a…›this line of inquiry
How do diffusion language models enable capabilities autoregressive architectures cannot achieve?
A broader line of inquiry — a family of 61 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 61
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- How can diffusion models predict future tokens without completing prior blocks?
- Do diffusion language models learn differently than autoregressive models?
- Can gradient-based control reach properties that autoregressive methods cannot?
- What structural differences between diffusion and autoregressive models enable bidirectional prompting?
- Why do autoregressive models fail at controlling syntactic structure and semantic content?
- Can diffusion language models match autoregressive inference speed in practice?
- How does selective looping in diffusion models differ from recurrence in autoregressive architectures?
- Can architecture changes and early stopping combine to close the diffusion inference gap?
- Can diffusion models perform infilling and reverse generation as naturally as forward generation?
- How does tokenization toward corpus mean affect downstream output diversity?
- Can fast-slow separation improve both memory and generation in language models?
- Does diffusion's control advantage come from speed gains or from architectural differences?
- Why do diffusion LLM answer tokens converge in confidence long before reasoning stabilizes?
- Can diffusion models condition on right context natively without special training for infilling?
- How do diffusion language models outpace autoregressive generation in speed?
- Why is reinforcement learning harder to apply to diffusion language models?
- How do autoregressive models constrain where chain-of-thought prompts can be positioned?
- What causes autoregressive generation to fail on out-of-corpus item identifiers?
- Can autoregressive models learn faithful translation to logical representations without semantic loss?
- Why does bidirectional attention in diffusion models prevent KV cache reuse?
- How does the discrete token bottleneck prevent gradient flow in language model control?
- Does iterative denoising order affect the reasoning style diffusion models learn?
- Why do diffusion models fail at inherently sequential problems?
- Can autoregressive models be trained to produce more cataphoric text?
- How does token-by-token generation constrain a model's ability to plan ahead?
- Could superposed decoding algorithms maintain multi-task representation during generation?
- How does merging retrieval and generation shift the computational bottleneck in dialogue systems?
- Why do hybrid paradigms outperform pure autoregressive or pure diffusion approaches?
- Can language models produce language more efficiently through interaction?
- Do bidirectional and any-order generation expose different parts of the joint distribution?
- How does the articulatory substrate explain direct speech-to-speech superiority over transcription pipelines?
- Can one streaming model handle turn-taking better than cascaded ASR-LLM-TTS?
- Can we measure how much prior errors bias subsequent token predictions?
- Can looped diffusion models outperform standard depth scaling at fixed parameters?
- Can feature disentanglement in gesture synthesis generalize to completely unseen voice distributions?
- How much does shared-prefix sampling reduce token redundancy empirically?
- Can accelerated sampling techniques from image generation speed up evolutionary search?
- How does smooth generation lead to proliferation without new viewpoints?
- How does causal multimodal modeling differ from encoder-decoder architectures?
- How do perception and generation share timing in a single causal stream?
- Do speech models learn the articulatory processes that produce acoustic signals?
- What makes seed data a bottleneck in synthetic generation pipelines?
- Why does token redundancy and poor readability emerge at trillion-parameter scale?
- Why does articulatory probing predict SSL model performance better than phonetic probing?
- Why does disrupting drafting flow improve synthesis of multiple sources?
- Why does the second loop do most of the productive refinement work?
- Does direct speech-to-speech generation really eliminate transcription latency?
- How much latency improvement comes from collapsing the speech pipeline?
- How does removing transcription change speech-to-speech generation latency?
- Can selective history filtering address topic drift that generation-time topic following cannot prevent?
- Can skipping transcription reduce speech dialogue latency below 300 milliseconds?
- What makes draft-centric systems better anchors for coherence than feed-forward outputs?
- How does cascading ASR, LLM, and TTS compound latency?
- How do early-prefix tokens control the generation of entire continuations?
- Does direct speech-to-speech generation actually reduce latency over ASR-LLM-TTS?
- Can articulatory inversion serve as a window into what speech models have learned?
- What makes session-aware multi-turn tracking necessary for asynchronous training?
- How do different speech encoder layers capture different types of gesture information?
- How does single-pass generation differ from multi-stage synthesis architecturally?
- Why do cascade pipelines fail to capture global motion structure?
- What temporal and spatial constraints does Space-Time U-Net solve?