icassp2027-experiment/manuscript/main/main.tex

% ICASSP 2027 content skeleton. The final paper must use the official 2027
% author kit and contain no visible \todo{} markers. Planned experiments are
% intentionally marked rather than presented as completed work.
\documentclass{article}
\usepackage{spconf}
\usepackage{amsmath,amssymb,graphicx,hyperref,xcolor}

\newcommand{\todo}[1]{\textcolor{red}{\textbf{[TODO: #1]}}}
\newcommand{\vect}[1]{\boldsymbol{#1}}
\setlength{\emergencystretch}{2em}

\title{Is It My Turn? Causal Local Evidence Adaptation of Omni Models for
Egocentric Floor Forecasting}

% TODO: Insert the final author list and affiliations after the ICASSP 2027
% review policy and author order are finalized.
\name{Internal Manuscript Draft}
\address{}

\begin{document}
\ninept
\maketitle

\begin{abstract}
Real-time audiovisual agents must decide not only what to say, but whether an
emerging floor opportunity belongs to them. This is difficult in multiparty
egocentric interaction: the target wearer is behind the camera, while the
visible people provide only indirect evidence about whether the surrounding
floor remains occupied. We study strictly causal forecasting of the wearer's
future speech activity every 200~ms. Although a general-purpose Omni model
provides rich long-context audio--visual representations, its global readout is
not explicitly trained to expose sub-second, wearer-relative floor evidence.
We therefore study a candidate causal local floor adapter that combines the global
Omni state with short-window mixed audio and an anonymous set of visible-person
observations, without assuming persistent speaker identity. On EasyCom,
\todo{report the final three-seed dense and committed-event
results: EOT-response and INT-yield recall, false-positive rate, latency,
confidence intervals, and oracle-gap recovery}. Controlled
\todo{time-shift/remove-vision/same-rate or capacity-matched} interventions
show that \todo{state the supported mechanism, without claiming addressee or
speaker identity}. These results demonstrate \todo{the narrow conclusion
actually supported by the final experiments}.
\end{abstract}

\begin{keywords}
conversational agents, turn-taking, egocentric video, multimodal adaptation,
voice activity projection
\end{keywords}

\section{Introduction}
\label{sec:intro}

Natural turn-taking requires a conversational system to anticipate not only
when speech will stop, but whether a floor opportunity actually belongs to it.
In a dyadic exchange, a pause after one interlocutor finishes often invites the
other to respond. In a group conversation, the same pause may instead mark a
handover between two other participants. An egocentric agent that reacts only
to silence or aggregate speech may therefore interrupt a conversation that
should continue without it. We study \emph{egocentric floor forecasting}:
continuously predicting the near-future speech participation of the
first-person wearer from strictly causal audio--visual observations.

Voice Activity Projection (VAP) introduced continuous prediction of future
voice activity for turn-taking \cite{ekstedt2022vap}. EgoSpeak brought this
formulation to causal, untrimmed first-person streams, using future camera-
wearer activity as an imitation target \cite{kim2025egospeak}. Complementary
work adds visual gaze, head pose, and facial behavior to predictive turn-taking
\cite{russell2025mmvap}, while MuVAP combines global acoustic dynamics with
visible face tracks to forecast visible floor holders in multiparty scenes
\cite{qi2026muvap}. Our target has a different role structure: the participant
whose future action is predicted is the one participant absent from the
visible set. The visible people are contextual evidence for the unseen
wearer's decision, not candidate output identities.

This role asymmetry exposes an adaptation problem for general-purpose Omni
models. Omni pretraining favors globally integrated representations useful for
multimodal understanding and generation. Egocentric floor control instead
depends on timely prosodic and visible-person activity at sub-second scale,
including whether other participants continue to occupy the floor. Increasing
model capacity or applying a generic parameter-efficient update does not
explicitly preserve or route this evidence to a dense wearer-relative
prediction head. The open question is therefore not merely whether an Omni
backbone can be applied to EgoSpeak, but what task structure is still required
to turn general audiovisual understanding into reliable floor control.

Our candidate decomposition separates long-context understanding from
short-window floor evidence. A
causal global Omni state represents the broader scene and conversation. In
parallel, a lightweight local path encodes recent mixed audio and a changing,
anonymous set of visible-person observations. A residual adapter makes this
local evidence accessible to the common ten-horizon activity head. The primary
variant deliberately avoids cross-time identity: the current development
oracle supports timely aggregate visible-other activity but not person
continuity. Track-aware structure is considered only if it defeats no-ID,
identity-corrupted, and capacity-matched controls.

The final paper will claim the following three contributions only if the
corresponding TODO gates pass:
\begin{itemize}
  \item We identify and test a temporal--role mismatch between a global Omni
  readout and causal egocentric floor forecasting for an invisible wearer.
  \item We evaluate a wearer-relative causal local floor adapter that combines
  long-context Omni information with short-window audio and anonymous
  visible-person evidence.
  \item We evaluate whether the adapter reduces erroneous wearer commits
  during multiparty floor transfers under an EOT/INT event protocol with
  explicitly annotated hard-negative regions.
\end{itemize}

\begin{figure}[t]
\centering
\fbox{\parbox[c][1.00in][c]{0.94\columnwidth}{\centering
\todo{Draw only the selected method: causal AV to the long-context Omni path;
the short-window mixed-audio and anonymous person-set path; residual fusion;
the common $10\times3$ head; and the separate EOT-response/INT-yield evaluation.
Mark physical cutoffs, trainable modules, person masks, and NULL/OFFSCREEN.
Do not draw oracle labels, rejected track variants, or multiple backbones.}}}
\caption{Long-context Omni understanding and short-window local floor evidence
for wearer-relative future-activity prediction.}
\label{fig:method}
\end{figure}

\section{Method}
\label{sec:method}

\subsection{Causal wearer-relative projection}

At decision time $t$, the model observes only the stream
$\mathcal X_{\leq t}$ and predicts $K=10$ future bins at $\Delta=0.2$~s. For
bin $k$, the multilabel target
$\vect y_{t,k}=[y^{\mathrm{bg}},y^{\mathrm{w}},y^{\mathrm{o}}]$ denotes
background, camera-wearer activity, and activity by one or more other people.
Wearer--other overlap is allowed. The model outputs
\begin{equation}
 \hat{\vect Y}_t=f_\theta(\mathcal X_{\leq t})
 \in[0,1]^{K\times3}.
 \label{eq:task}
\end{equation}
The wearer-activity forecast also provides a causal control score. When the
wearer is silent, $\hat y^{\mathrm w}_{t,1}$ acts as a near-term take-turn
score; while the wearer is active, $1-\hat y^{\mathrm w}_{t,1}$ acts as a
yield score. These scores are trained only from future activity targets.
Committed EOT-response and INT-yield events used for behavioral evaluation are
defined separately from participant-resolved annotations in
Sec.~\ref{sec:experiments}.

\subsection{Global Omni baseline}

Let $\vect g_t\in\mathbb R^{d_g}$ be the causal global readout of
Qwen2.5-Omni at the physical cutoff $t$ \cite{xu2025qwenomni}. The frozen
baseline trains only a dimension-matched prediction head. A matched PEFT/LoRA
baseline updates a pre-registered subset of the Omni model under the same
examples, context, optimizer budget, head, and evaluator. Its purpose is to
test whether generic parameter adaptation is sufficient; one negative recipe
is not interpreted as a general limitation of LoRA or all Omni models.

\todo{Freeze the exact Qwen2.5-Omni-3B revision, processor, input rates,
streaming/cache semantics, readout layer, context, dtype, LoRA targets/rank,
trainable parameter budget, and cutoff audit. Add Qwen2.5-Omni-7B only as a
capacity-scaling control. Do not add another backbone family unless it is
needed to support and experimentally earns a cross-family claim.}

\subsection{Causal local floor adapter}

The candidate local path operates on the most recent $L_\ell$ seconds. At local instant
$\tau\leq t$, a causal audio encoder produces $\vect a_\tau$ from mixed audio.
For the anonymous visible-person set
$\mathcal P_\tau=\{(\vect x_{\tau,i},\vect b_{\tau,i})\}_{i=1}^{N_\tau}$,
a shared visual encoder and masked set operator produce
\begin{equation}
 \vect v_\tau=\mathrm{SetPool}
 \left(\{E_v(\vect x_{\tau,i},\vect b_{\tau,i})\}_{i=1}^{N_\tau}
 \cup\{\vect v_\varnothing\}\right),
 \label{eq:set}
\end{equation}
where $\vect v_\varnothing$ handles an empty or offscreen visible set. The
operator is permutation invariant and uses no participant-ID embedding or
cross-time identity. A causal short-window encoder summarizes local dynamics,
\begin{equation}
 \vect \ell_t=F_\ell
 \left(\{[\vect a_\tau;\vect v_\tau]\}_{\tau\in(t-L_\ell,t]}\right).
 \label{eq:local}
\end{equation}
The adapter preserves the global initialization through a zero-initialized
residual output,
\begin{equation}
 \vect z_t=\vect g_t+W_o\,
 \sigma\!\left(W_g\vect g_t+W_\ell\vect\ell_t\right),
 \qquad W_o\vert_{\mathrm{init}}=0,
 \label{eq:fusion}
\end{equation}
and the common head maps $\vect z_t$ to Eq.~\eqref{eq:task}. This implements an
evidence-access hypothesis rather than adding an unconstrained second model:
the global path retains broad context, while the local path preserves timely
floor evidence.

\todo{Freeze $L_\ell$, local audio/video rates, crop source and resolution,
set pooling, causal temporal encoder, masks, reset rules, parameter count, and
latency. If the local path runs faster than the 5-Hz global/output grid, compare
it with a same-parameter 5-Hz variant before using the term ``dual-rate.'' Use
provided causal boxes only as a clearly labeled method-isolation condition;
online detection/tracking is not a paper contribution unless actually run.}

\subsection{Training}

The primary objective is class-balanced binary cross entropy over valid
horizons and activity channels,
\begin{equation}
 \mathcal L_{\mathrm{act}}=
 \sum_{t,k,c}m_{t,k}w_c\,
 \mathrm{BCE}(\hat y^c_{t,k},y^c_{t,k}).
 \label{eq:loss}
\end{equation}
No transcript, target-of-speech label, close microphone, future frame, or
per-person speech label enters the deployable model. \todo{Add an active-speaker
or event auxiliary loss only if it improves the frozen primary protocol and
its label/input boundary is fully reported.}


\section{Experiments}
\label{sec:experiments}

\subsection{Dataset and Data Preparation}

We evaluate on EasyCom following the EgoSpeak split, using Sessions 4--12 as
the training/development pool and Sessions 1--3 only for final evaluation
\cite{donley2021easycom}. The participant-resolved speech annotations are
converted into two parallel representations. First, they are downsampled to
5~Hz to construct the EgoSpeak-compatible ten-horizon targets for background,
camera-wearer, and aggregate non-wearer activity. Second, the original
participant timeline is retained to derive committed interaction events for
behavioral evaluation.

We consider two wearer-relative events. An \emph{EOT-response} denotes a
non-wearer floor release followed by the wearer taking the floor, whereas an
\emph{INT-yield} denotes a non-wearer takeover while the wearer is speaking,
followed by the wearer relinquishing the floor. Candidate events are generated
automatically from participant-level onset/offset annotations. Mid-turn pauses,
other-to-other handovers, backchannels, and non-takeover overlaps are retained
as hard-negative regions rather than positive events. Annotation-corrupted or
unresolvable boundaries are excluded from scoring. Ambiguous candidates are
manually adjudicated.

Participant identity is used only to construct these evaluation annotations
and is never provided to the prediction model. Provided person boxes are used
only for the causal visible-person input; transcripts, target-of-speech labels,
close-microphone audio, and future observations are unavailable to the model.
\todo{Report duration and event statistics: EOT positives/hard negatives,
INT positives/hard negatives, excluded candidates, and manual-adjudication
count for train/development and test. Freeze the event-construction and
adjudication protocol before reporting semantic EOT/INT results; until then,
retain the timing-derived proxy terminology.}

\subsection{Model Configuration and Training}

Our main backbone is Qwen2.5-Omni-3B. We compare a frozen global Omni readout,
matched PEFT/LoRA adaptation, global features with recent local audio, a
parameter- and history-matched temporal local branch, and the proposed causal
local audio--visual adapter. Qwen2.5-Omni-7B is used as a capacity-scaling
control. All learned variants use identical causal examples, prediction
horizons, prediction heads, and the class-balanced binary cross-entropy
objective in Eq.~\eqref{eq:loss}.

Model selection is performed using training-side data only. Event operating
points are selected from session-level out-of-fold predictions over
Sessions~4--12; the final model is then retrained on all nine sessions before
evaluation on Sessions~1--3. Final results are averaged over three random
seeds.
\todo{Freeze context length, local window/rates, optimizer, epochs, LoRA
configuration, trainable parameter counts, and inference latency.}

\subsection{Evaluation Protocol}

We first retain dense future-activity evaluation for direct comparison with
EgoSpeak, reporting AP at the 0.2--2.0~s horizons and their mean mAP.

Behavioral evaluation follows a committed-event protocol. For both
EOT-response and INT-yield, each gold event is matched to at most one committed
prediction and is scored by recall and signed detection latency. A prediction
inside a task-specific hard-negative region is counted as a false positive.
EOT hard negatives include mid-turn continuations and competing
other-to-other floor transfers; INT hard negatives include backchannels and
overlaps in which the wearer retains the floor. Regions excluded during
adjudication do not contribute to either the positive or negative denominator.

We report the recall--false-positive trade-off and a common operating point at
matched recall $R=0.8$. The primary behavioral quantities are therefore EOT
false-positive rate, INT false-positive rate, and signed latency at the frozen
operating point. We additionally report performance on the other-to-other
subset to test whether local visual evidence specifically reduces erroneous
wearer turn-taking during multiparty floor transfers. \todo{Freeze the FPR
denominators and report signed median latency separately for EOT and INT.}


\section{Results and Analysis}
\label{sec:results}

\subsection{Main Comparison}

Table~\ref{tab:main} compares the proposed causal local audiovisual adapter
with an EgoSpeak reproduction, the frozen Qwen2.5-Omni-3B global state $G$,
matched PEFT, $G$ plus recent local audio, and a parameter- and history-matched
temporal local branch. Qwen2.5-Omni-7B is included only as a capacity-scaling
control. All methods use the same split, causal inputs, targets, and evaluation
protocol. Event thresholds are selected on training-side calibration data and
frozen before test evaluation. \todo{Freeze all baseline implementations and
verify identical targets, causal validity, and evaluation.}

\begin{table}[t]
\caption{Confirmatory EasyCom results. EOT and INT false-positive rates are
measured at matched recall $R=0.8$. E/I Lat. gives signed median EOT/INT
latency in seconds; values closer to zero are better.}
\label{tab:main}
\centering
\small
\resizebox{\columnwidth}{!}{%
\begin{tabular}{lccccc}
\hline
Method & 0.2-s mAP $\uparrow$ & Mean mAP $\uparrow$
& EOT FPR@$R=.8$ $\downarrow$ & INT FPR@$R=.8$ $\downarrow$ & E/I Lat.\\
\hline
EgoSpeak reproduction       & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
Omni-3B frozen $G$          & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
Omni-3B matched PEFT        & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
$G+$local audio             & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
$G+$matched temporal local  & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
$G+$causal local AV (ours)  & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
Omni-7B scaling control     & \todo{--} & \todo{--} & \todo{--} & \todo{--} & \todo{--/--}\\
\hline
\end{tabular}}
\end{table}

\textbf{Dense forecasting.}
The proposed local AV adapter improves 0.2-s mAP from
\todo{$X$ to $Y$} over the strongest matched baseline and improves
mean mAP by \todo{$Z$ points}.
The gain is largest at \todo{short prediction horizons / transition-adjacent
frames / specify actual observation}, indicating that local audiovisual
evidence primarily improves \todo{the interpretation supported by the data}.

\textbf{Committed behavior.}
At matched EOT-response recall $R=0.8$, the proposed adapter reduces EOT FPR
from \todo{$X$} to \todo{$Y$}; at matched INT-yield recall, it changes INT FPR
from \todo{$A$} to \todo{$B$}. Its signed median EOT/INT latencies are
\todo{$+0.XX/+0.XX$~s}. These recall-matched error rates and signed latencies
test whether fewer erroneous commits arise from better floor decisions rather
than indefinite waiting.

\textbf{Comparison with adaptation baselines.}
Matched PEFT yields \todo{result}, whereas adding recent local audio alone
yields \todo{result}.
The parameter- and history-matched temporal branch achieves
\todo{result}.
The remaining gap between this control and the proposed adapter therefore
suggests that \todo{correctly timed audiovisual local evidence contributes
beyond additional temporal capacity}.
\todo{If this difference is not statistically supported, state instead that
the local temporal branch accounts for most of the gain.}

\subsection{Information Opportunity}

Before designing the deployable adapter, we measured whether local floor
information was useful at all using an analysis-only oracle.
The oracle receives perfectly timed visible-other speech activity through a
non-deployable pathway and is never used in the main comparison.

On development Session~12, aggregate visible activity improved 0.2-s mAP by
6.97 points and ten-horizon mean mAP by 2.91 points over the matched
global-plus-local-audio baseline.
In contrast, visibility alone and one-second-delayed activity remained close
to the audio baseline, while person-specific histories did not outperform
identity-shuffled histories.

These controls indicate that the useful signal is primarily
\emph{timely local floor evidence}, rather than participant identity or
static visibility.
The proposed adapter is therefore designed to approximate this information
using only causally available audiovisual observations.

\todo{After completing all seeds, report mean $\pm$ confidence interval here.
If the oracle effect is unstable, describe it qualitatively and omit the exact
numbers from the four-page version.}

\subsection{Mechanism Analysis}

To determine whether the gain arises from correctly aligned visual floor
evidence rather than additional parameters or generic visual context, we
evaluate the controls in Table~\ref{tab:ablation}.

\begin{table}[t]
\caption{Mechanism controls for the selected Omni-3B local adapter. All
variants use the same prediction head and comparable temporal capacity; FPRs
are measured at matched recall $R=0.8$.}
\label{tab:ablation}
\centering
\small
\resizebox{\columnwidth}{!}{%
\begin{tabular}{lcccc}
\hline
Local condition & Mean mAP $\uparrow$ & EOT FPR $\downarrow$
& O$\rightarrow$O FPR $\downarrow$ & INT FPR $\downarrow$\\
\hline
Remove vision           & \todo{--} & \todo{--} & \todo{--} & \todo{--}\\
Time-shift vision       & \todo{--} & \todo{--} & \todo{--} & \todo{--}\\
Full-frame vision       & \todo{--} & \todo{--} & \todo{--} & \todo{--}\\
Correct local AV        & \todo{--} & \todo{--} & \todo{--} & \todo{--}\\
\hline
\end{tabular}}
\end{table}

Removing local vision changes \todo{EOT/O$\rightarrow$O/INT FPR by $X/Y/Z$},
while shifting the visual stream by \todo{$\Delta t$} produces \todo{result}.
Replacing person-local evidence with full-frame vision produces \todo{result}.
If aligned local AV improves estimates of surrounding floor occupancy, its
largest selective reduction should appear on the O$\rightarrow$O hard-negative
subset. Together, these comparisons \todo{support / do not support} that
mechanism beyond generic visual context.

\todo{Report paired intervals and the supported interpretation. If correct
and shifted/removed vision do not separate, drop the audiovisual
evidence-routing claim. If dense mAP improves but event behavior does not, do
not claim improved floor control. If legitimate onset or yield degrades, do
not describe lower false entry as success.}

\subsection{Scaling Control}

Increasing the frozen Omni backbone from 3B to 7B yields
\todo{$X$ points} in 0.2-s mAP and \todo{EOT/INT FPR of $Y/Z$} at matched
recall.
The proposed 3B adapter \todo{remains better than / approaches / is surpassed
by} the 7B scaling control.

This result suggests that \todo{architectural access to local causal evidence
provides a complementary benefit to backbone scaling}.
Because both backbones belong to the Qwen2.5-Omni family, we do not claim that
this conclusion generalizes to all omni-modal models.

\subsection{Limitations and Qualitative Analysis}

The results support \todo{precise supported claim}, but several limitations
remain.
First, \todo{EasyCom size/session diversity limitation}.
Second, the adapter relies on \todo{state the actual visible-person box or
tracking assumption}.
Third, \todo{failure condition: occlusion, overlapping speakers, rapid floor
exchange, etc.} remains challenging.

\todo{Add one compact, pre-selected qualitative timeline only if space permits:
one multi-other false-entry improvement paired with one preserved real wearer
onset or one failure. Examples illustrate aggregate rates and never establish
the claim by themselves.}

\section{Conclusion}
\label{sec:conclusion}

This study asks whether a general-purpose Omni representation can be adapted
for causal floor forecasting of an invisible egocentric wearer. The candidate
method separates long-context Omni understanding from short-window local floor
evidence without assuming persistent speaker identity. \todo{State the final
learned-method improvement, matched-recall EOT/INT false-positive rates,
signed latency, causal-control evidence, and main limitation. If no learned
variant passes these tests, reframe the submission as an information audit or
do not make the method claim.}

\bibliographystyle{IEEEbib}
\bibliography{references}

\end{document}