Read-Best Is Not Steer-Best: A Probing--Steering Layer Dissociation in Omni-Modal Large Language Models

Yibo Wang; Jisheng Dang; Bimei Wang; Yitao Wu; Wencan Zhang; Hong Peng; Jizhao Liu; Bin Hu

cs.CLAug 24, 2026

Read-Best Is Not Steer-Best: A Probing--Steering Layer Dissociation in Omni-Modal Large Language Models

Authors: Yibo Wang, Jisheng Dang, Bimei Wang, Yitao Wu, Wencan Zhang, Hong Peng, Jizhao Liu, Bin Hu,

Organizations: Lanzhou University, Lanzhou, China · Hainan University, Haikou, China · School of Computing, National University of Singapore, Singapore · Cloud and AI BU, Huawei, Shenzhen, Guangdong 518129, China

Abstract

Omni-modal large language models integrate text, audio, and image signals into a shared residual stream, where concepts such as emotion can be linearly decoded and causally modified by activation steering. A common but rarely tested assumption is that the layer with the highest probing accuracy is also the best layer for steering, so injection layers are often selected by probe performance. We provide the first causal test of this assumption across three independently developed omni-modal models and find that it fails. Reading and intervention rely on different layers, a phenomenon we call the probing-steering layer dissociation. Using emotion as a controlled testbed, we measure layer-wise readability and steerability across text, audio, and image inputs. Probe-best layers vary widely across architectures, while steering-effective layers consistently fall within a narrow mid-to-late range of normalized depth. Paired random-direction controls show an approximately 26-fold causal gap, ruling out random perturbation and direction quality as explanations. Logit-lens analysis reveals a staged forward process: causal handle, probing saturation, and vocabulary commitment, and motivates a two-factor account in which steering effectiveness depends on both representational readability and downstream plasticity. These results show that probing accuracy is a poor heuristic for selecting intervention layers and suggest a cross-architecture mid-to-late selection criterion. We also identify a cross-modal emotion subspace organized by valence and arousal, with joy acting as a stable anchor across models. Code and data: https://github.com/YiboWang2002/Read-Best-Is-Not-Steer-Best.

Figures & tables

Fig. 1 : Read-best is not steer-best. In an omni-modal LLM, probe readability (blue) peaks at a late layer while the steering effect of an emotion direction (red) peaks earlier, in the mid-to-late zone. Injecting at the read-best layer is about $26\times$ weaker than in the steering zone. Curves are schematic. Full results in Figure 3 and §5.

Fig. 2 : Overview of the study. Text, audio, and image inputs enter an omni-modal LLM through a shared residual stream. A linear probe reads a concept best at a late layer (probe-best), while steering an emotion direction is most effective at a mid-to-late layer (steer-best), a controlled gap of about $26\times$ . A logit-lens analysis orders the forward pass as causal handle, probing saturation, then vocabulary commitment, placing the steering window before commitment and motivating the two-factor model $H_{\mathrm{thresh}}$ . Bottom: (A) the steer-best layer clusters in a shared band across three architectures while the probe-best layer scatters; (B) the logit-lens mechanism; (C) controls and external judges (VAD-BERT, GPT-5.5).

Fig. 3 : The probing–steering layer dissociation at a glance. (a) On Qwen2.5-Omni-7B the probe-accuracy peak (L27) and the steering-effect peak (L20) are 7 layers apart, with the band marking the mid-to-late steering zone [L14, L22]. (b) Across three architectures the steer-best layer stays in a narrow band of normalized depth

[0.48,0.74]

, while the probe-best layer scatters across

[0.29,1.00]

(MiniCPM shows a reversed ordering). Full evidence in §5–§6.

Fig. 4 : Cross-modal emotion geometry in Qwen2.5-Omni-7B. Whitened M3 activations at L14 (375 samples) projected onto the supervised valence and arousal axes. Color encodes emotion, marker shape encodes modality, and shaded regions are 95% confidence ellipses. The five emotions fall at their expected circumplex positions and the three modalities of each emotion overlap, visualizing the dimensional sharing reported in the Supplementary Material.

Condition	$\Delta_{\text{joy}}$	95% CI	VADER $\Delta$	Random $\Delta$
Ours (L10/14/18/22)	+0.549	$[+0.393,+0.693]$	+0.201	$-0.098$
Baseline-A (centroid)	+0.573	$[+0.437,+0.707]$	+0.242	$-0.098$
Baseline-B (L18)	+0.152	$[-0.004,+0.313]$	+0.143	$-0.067$
Baseline-C (L27)	+0.021	$[-0.078,+0.128]$	0.000	+0.019

TABLE I : Four-condition joy steering on Qwen2.5-Omni-7B. All conditions use

\alpha=+0.12

, the same prompts (

n=30

) and direction protocol, differing only in layer selection. Ours and Baseline-B/C use the raw_cmd_pc_orth_k2 direction, Baseline-A the centroid.

\Delta_{\text{joy}}

is the VAD-BERT change of steered relative to baseline (95% CI from 1000 bootstrap), Random

\Delta

is the same-condition random control, and the OOD rate is 0% throughout. Visual comparison in Figure 5 .

Fig. 5 : Four-condition joy steering effect. Solid bars are

\Delta_{\text{joy}}

of the main direction, lighter bars the same-condition random control, and error bars are 95% bootstrap CIs. Multi-layer injection (Ours, Baseline-A) rises sharply, while single-layer probe-best L27 (Baseline-C) sits near zero and matches its random control. The roughly 26-fold gap is the empirical signature of the dissociation. Exact values in Table I .

Emotion	Single-layer best	$\Delta_{\text{target}}$
joy	L18	+0.194
anger	L17	+0.079
sadness	L20	+0.040

TABLE II : Single-layer steering best layer and effect over the 28 layers of Qwen2.5-Omni-7B (

\alpha=+0.12

n=20

prompts

\times

5 seeds, VAD-BERT). All three emotions peak in the mid-to-late range (L17–L20), while the end layers are near 0. Multi-layer effect in Table I .

Fig. 6 : The 28-layer dissociation curves of Qwen2.5-Omni-7B. Blue is layer-wise probing accuracy (left axis), and the colored lines are the single-layer steering effect for joy / anger / sadness (right axis,

\alpha=+0.12

). The probing peak (L27) and steering peak (L20) are 7 layers apart, and the shading marks the mid-to-late steering zone [L14, L22].

Model	Layers	Probe-best	Steer-best	Gap	$r$
Qwen	28	L27 (1.00)	L20 (0.74)	7	+0.452
Phi-4	32	L25 (0.81)	L15 (0.48)	10	+0.546
MiniCPM	36	L10 (0.29)	L19 (0.54)	9	+0.362

TABLE III : Cross-architecture probing–steering dissociation across three models (Qwen2.5-Omni, Phi-4-Multimodal, MiniCPM-o-4.5). Probe and steer peaks are measured per model by a layer-wise scan (5 seeds, VAD-BERT); each cell gives the layer and its normalized depth (layer /

L_{\max}

). Under the pre-registered criterion (

\lvert\text{peak gap}\rvert\geq 5

layers), the gap passes in all three (7 / 10 / 9 layers), and the steering peak (0.48–0.74) is far more concentrated than the probing peak (0.29–1.00).

Fig. 7 : Cross-architecture probing–steering dissociation (signature figure). Subplots are Qwen2.5-Omni, Phi-4-Multimodal, and MiniCPM-o-4.5. The probe-best layer (triangle) drifts across normalized depth 0.29–1.00, nearly the full network, while the steer-best layer (star) stays in the steering-effective zone [0.48, 0.74] (band) across all three architectures.

Hypothesis	Statement	Verdict
H1	Vocabulary commitment precedes probing saturation (L23 $<$ L20)	$\times$
H2	The steering peak is adjacent to the commitment window ( $\lvert\text{L19}-\text{L23}\rvert\leq 4$ )	$\checkmark$
H3	Probing saturation precedes vocabulary commitment (L20 $<$ L23)	$\checkmark$

TABLE IV : Adjudication of the three pre-registered logit-lens hypotheses (Qwen2.5-Omni-7B, 90 emotion-conditioned trajectories; layer numbers are medians of the timing).

Fig. 8 : Cross-layer logit-lens trajectory of the emotion token (joy / anger / sadness, median rank over

n=30

prompts). The vertical axis is the median vocabulary rank of the target emotion token (log scale, inverted, so the top is rank 1). Shaded bands mark the three landmark layers: steering peak (L19), probing saturation (L20), and vocabulary commitment (L23).

Fig. 9 :

H_{\mathrm{thresh}}

prediction versus the measured steering effect across models. Solid lines are measured, dashed lines the

H_{\mathrm{thresh}}

prediction, stars the measured peak, and triangles the predicted peak. Qwen is hit exactly (error 0), Phi-4 is off by 1 layer, and MiniCPM by 9 layers (red shading marks the gap). Fit parameters in the Supplementary Material.

Fig. 10 : Qualitative cross-modal joy steering. The same multi-layer joy injection (L10/14/18/22, $\alpha=+0.12$ ) is applied to an image input (EmoSet, anger) and an audio input (RAVDESS, sad speech). The baseline generation conveys the original negative tone, while the joy-steered generation shifts to a celebratory one, with the VAD-BERT joy probability rising from $0.00$ to $0.99$ in both cases. Quantitative results in Table V .

Input context	$\Delta_{\text{joy}}$	$\Delta_{\text{sadness}}$	$\Delta_{\text{anger}}$
text-only	+0.549	+0.101	+0.166
audio-context	+0.670	+0.361	+0.282
image-context	+0.334	+0.162	+0.179

TABLE V : Multi-layer emotion steering under three input contexts (VAD-BERT

\Delta_{\text{target}}

\alpha=+0.12

; text 30 prompts, audio 100 RAVDESS, image 102 EmoSet). joy is strongly positive under all three modalities, the OOD rate is 0, and random controls are near zero.

Model	Injection layers	$\Delta_{\text{joy}}$	95% CI
Qwen2.5-Omni	L10/14/18/22	+0.549	$[+0.393,+0.693]$
Phi-4-Multimodal	L12/16/20/24	+0.415	$[+0.289,+0.536]$
MiniCPM-o-4.5	L14/18/22/27	+0.022	$[-0.064,+0.122]$

TABLE VI : Cross-model replication of joy steering (VAD-BERT

\Delta_{\text{joy}}

\alpha=+0.12

, layers mapped by normalized depth). Qwen and Phi-4 replicate a strong effect (CI excludes 0); MiniCPM-o is null (CI contains 0), matching its boundary-condition diagnostics in §4 / §5.5.

Language Model Steering

Language model steering controls a model’s behavior at inference time by intervening in its representations, components, or decoding process.

See all 283 papers

Some Emotions Run Deeper: Layer-wise Probing and Causal Intervention in Large Language Models
Sep 1, 2026Tian Fang, Gaël Guibon, Davide BuscaldiCausal Interventions in Language Models Emotion Representation in Language Models
1Université Sorbonne Paris Nord, CNRS, Laboratoire d’Informatique de Paris Nord, LIPN · 2LORIA, CNRS, Université de Lorraine
Steer Like the LLM: Activation Steering that Mimics Prompting
May 5, 2026Geert Heyman, Frederik VandeputteLanguage Model Steering LLM Prompting
Nokia Bell Labs, Belgium.
Deployable Per-Instance Multi-Layer Activation Steering for Large Language Models
Aug 9, 2026Muhammad Faishal Adly Nelwan, Alfan Farizki WicaksonoLanguage Model Steering Controllable Text Generation
Faculty of Computer Science Universitas Indonesia

Sep 1, 2026J/K move · Enter open · S save

Some Emotions Run Deeper: Layer-wise Probing and Causal Intervention in Large Language Models

Tian Fang, Gaël Guibon, Davide Buscaldi

1Université Sorbonne Paris Nord, CNRS, Laboratoire d’Informatique de Paris Nord, LIPN · 2LORIA, CNRS, Université de Lorraine

Causal Interventions in Language Models Emotion Representation in Language Models Language Model Probing

Open paper AskPDF arXiv

Emotion is expressed in text along a wide spectrum, from surface lexical cues to inferences entangled with content. Most layer-wise analyses of emotion in LLMs use a single corpus, leaving open whether the depth at which emotion becomes accessible is a property of the model or also of the text source. We investigate this across three datasets spanning different degrees of explicitness and contextualization in emotion expression (Twitter posts, Reddit comments, and autobiographical narratives) and eight 1B--9B open-weight LLMs from the Llama, Qwen, and Granite families. We combine layer-wise probing with offline feature scaling and online forward interventions, transfer analyses, and an early-exit classifier. We find that (i) the best probing layer shifts systematically across corpora, from input-adjacent layers to over half model depth, and this ordering persists after matching label-by-length-bin distributions; (ii) across the evaluated settings, forward-pass interventions on probe-selected bands reduce test accuracy by 5--6 points more than same-width random bands (

q < 0.01

); (iii) selected bands transfer across datasets and emotion categories, suggesting partially shared affective information rather than strictly per-emotion substrates; and (iv) probe-selected early-exit representations outperform full-depth exits by

6.9

percentage points on average.

Abstract

Language Model Steering

Explore similar work

Some Emotions Run Deeper: Layer-wise Probing and Causal Intervention in Large Language Models

Steer Like the LLM: Activation Steering that Mimics Prompting

Deployable Per-Instance Multi-Layer Activation Steering for Large Language Models