encoder-projector-decoder-audio-language-models-speaking-style-linearly-decodable-late-audio-encoder CLM-0382In encoder-projector-decoder audio-language models, speaking style is linearly decodable from late audio-encoder layers at about 82-85% on a 7-way style task, yet open-ended tone prediction at the output reaches only about 20-54%, and which end of that range a model lands on tracks its training objective rather than its encoder: a model trained to generate speech keeps acoustic sensitivity through the decoder while text-only-trained siblings with a nearly identical encoder answer from the words instead.
Capability: Understanding speech and audio
Observed on
Four open models (Whisper-large-v2, Qwen2-Audio-7B-Instruct, Qwen2.5-Omni-7B, Chroma-4B) on the Expresso dataset with fixed text and 7 speaking styles, four speakers, mean-pooled r.
Sources
- Measured: layerwise linear probes with leave-one-speaker-out, CKA, per-style open-ended tone accuracy, and an NMI-based content/style leakage ratio (0.977 Qwen2-Audio vs 0.272 Omni). The Qwen2-Audio/Omni pair holds encoder roughly fixed and varies objective, which is the strongest part of the design; only one model per objective type, four speakers, one dataset, and the leakage ratio's absolute values are entropy-biased.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- In audio language models that feed encoder outputs into a text LLM, downstream understanding accuracy across speech, sound and music is set by how semantically text-aligned the encoder is rather than by whether its representation is discrete or continuous — reconstruction-oriented codecs with high acoustic fidelity underperform, semantically distilled tokenizers can beat their own continuous teacher, and scaling the LLM backbone does not recover the lost information.Understanding speech and audio · reviewed by AI
- On multi-session dialogue questions whose answers depend only on how an utterance was delivered, a cascaded ASR-then-LLM pipeline (Whisper large-v3 into a frontier model) scores below 7B audio-native models on the same clips, and below even a transcript-only baseline, because the transcription step drops delivery cues entirely and adds content errors.Understanding speech and audio · reviewed by AI
- In LLM-based ASR, steering decoding toward acoustic evidence only helps if candidates are first restricted to tokens within a small likelihood margin of the greedy token: unconstrained selection by an acoustic compatibility score removes hallucinated outputs but destroys transcription accuracy through unbounded insertions, while the constrained version removes a large share of hallucinations at near-zero WER/CER cost.Stating false facts confidently · reviewed by AI
- For spoken language models answering questions over long dialogue histories, converting the history to transcripts beats keeping it as audio when the answer is a lexical detail (exact slot values across dense turns: mean 71.3% vs 42.6%), but loses when the answer depends on how something was said, where audio history wins for six of seven models and by a larger margin once lexical emotion cues are removed.Understanding speech and audio · reviewed by AI
- For large audio language models on multiple-choice audio reasoning benchmarks (MMAR, MMAU-mini), accuracy does not track whether the model's chain of thought is acoustically grounded or causally drives the answer: several models answer correctly from perception steps that assert unsupported labels, and for some models replacing the reasoning chain with vacuous or fabricated content leaves the final answer unchanged.Understanding speech and audio · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: An audio-language model trained only for text prediction whose output tone accuracy matches its late-encoder probe accuracy, or a speech-generating model whose decoder shows the same content-driven leakage ratio as its text-only counterpart with the same encoder.