Instruction-Tuned, but Not More Verifiable Instruction-Following: A Cross-Task Diagnosis for LoRA Adapters

Junyi Zou

cs.LGMar 23, 2026

Instruction-Tuned, but Not More Verifiable Instruction-Following: A Cross-Task Diagnosis for LoRA Adapters

Authors: Junyi Zou

Organizations: Zjydiary Group

Abstract

Adapters are often selected and deployed based on nominal labels (e.g., instruction-tuned), which implicitly suggest what capability improves after adaptation. We test whether nominal training objectives reliably align with realized cross-task capability gains by evaluating the same LoRA adapter across tasks. Our strongest evidence is tied to strict, automatically verifiable instruction following as measured by IFEval: across multiple seeds, base models, and LoRA settings, nominal labels recurrently but not universally fail to predict improvements on this verifiable target, with clear configuration sensitivity including a near-zero or negative case. As an illustrative strongest-case example in a controlled instruction-versus-numeric setting, an instruction-tuned adapter substantially improves off-target NM-based numeric benchmark performance from 0.133 to 0.632 while not improving verifiable instruction following on IFEval (ILA: 0.313 to 0.271; PLA: 0.250 to 0.143; values rounded to three decimals). We refer to this nominal-versus-realized mismatch pattern as capability drift as a descriptive label. The mismatch is visible in the raw cross-task performance matrix; we use a drift score only as a compact summary in the same units as the underlying metrics, not as a new formal metric contribution. Evidence from broader instruction-following benchmarks is benchmark-dependent and mixed, reflecting heterogeneity in how instruction following is operationalized; we therefore do not treat cross-benchmark agreement as a premise. Overall, the practical takeaway is to perform routine cross-task evaluation before deployment and to avoid treating nominal labels as reliable capability proxies.

Figures & tables

Model	Nominal task	Numeric NM	IFEval ILA	IFEval PLA	Realized change
Base	–	0.133	0.313	0.250	Reference point.
Reason adapter	Numeric	0.309	0.271	0.179	Numeric gain aligned with target.
Instruction adapter	Instruction	0.632	0.271	0.143	Numeric gain, no IFEval gain.

Table 1: Illustrative strongest-case cross-task evaluation (main setting). We report NM-based numeric benchmark performance (NM) and IFEval strict, automatically verifiable instruction following (ILA/PLA), with values rounded to three decimals. This example provides a concrete view of the mismatch in raw cross-task metrics; robustness results are summarized in Table 2 and Appendix Table 6 .

Figure 1: Drift score summary. Heatmap of $\mathrm{DriftScore}(a\rightarrow b)$ computed from cross-task evaluation metrics. Compact overview of score magnitudes from the same evaluations; see § 5 for context.

Panel of Figure 2 — Figure 2: Robustness quadrant plot (primary evidence). Per-run target gain on IFEval PLA (instr $-$ base) versus off-target gain on numeric NM (instr $-$ base) for the robustness evaluations underlying Table 2 and Appendix Table 6 . Most points lie above $y{=}x$ (positive drift), with occasional near-zero or negative cases.

Slice	Model	LoRA setting	$n$	Drift score ( $\uparrow$ )
Seeds	qwen3_8b	r16_attnmlp_do005_lrmain	5	0.511 $\pm$ 0.178
Seeds	qwen3_8b	r8_attnmlp_do005_lrmain	5	0.495 $\pm$ 0.150
Models	qwen3_14b	r16_attnmlp_do005_lrmain	2	0.667 $\pm$ 0.014
Models	qwen3_8b	r16_attnmlp_do005_lrmain	2	0.597 $\pm$ 0.010
Models	llama31_8b_instruct	r16_attnmlp_do005_lrmain	2	0.394 $\pm$ 0.016
Models	qwen25_7b_instruct	r16_attnonly_do0_lrsmall	1	-0.040

Table 2: Robustness summary of drift score (instr

\rightarrow

numeric reasoning). Drift score remains positive across five seeds (Qwen3-8B) and across multiple base models and LoRA settings, with one model/setting showing near-zero or slightly negative drift (configuration sensitivity).

Panel of Figure 3 — Figure 3: Selective shifts in verifiable instruction following (IFEval). Category-level and type-level shifts for the instruction adapter relative to base on IFEval. This breakdown is qualitative; low-support types should be interpreted cautiously.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

	Numeric EM	Numeric NM	IFEval ILA	IFEval PLA
base	0.000	0.133	0.312	0.250
reason	0.042	0.306	0.271	0.179
instr	0.000	0.640	0.271	0.143
domain	0.000	0.281	0.345	0.148

Appendix

Table 3: Extended cross-task evaluation including a domain adapter (secondary). This table is provided for completeness and is not central to the paper’s primary robustness claim, which is established by Table 2 and Appendix Table 6 ; Table 1 is illustrative. All values are reported as fractions in

[0,1]

; numeric benchmark performance is reported using both exact match (EM) and numeric match (NM).

Target metric	OffTargetGain (NM)	TargetGain (IFEval)	DriftScore (instr $\rightarrow$ numeric reasoning)
PLA	0.499	$-0.107$	0.606
ILA	0.499	$-0.042$	0.541

Appendix

Table 4: Target-metric sensitivity check (PLA vs ILA). Instruction

\rightarrow

numeric reasoning drift score recomputed from Table 1 using IFEval PLA or ILA as the instruction target metric.

Model	IFBench PLA (strict) base	IFBench PLA (strict) instr	FollowBench rule mean base	FollowBench rule mean instr
Qwen3-8B	0.122	0.112	0.000	4.000
Qwen3-14B	0.143	0.116	0.000	2.000
Llama-3.1-8B-Instruct	0.139	0.119	4.000	4.500

Appendix

Table 5: Supplementary multi-benchmark instruction following. FollowBench rule-based mean and IFBench strict PLA for base vs instruction adapter. Values are computed from benchmark-suite outputs. FollowBench, IFBench, and IFEval differ in scale and emphasis, so absolute values should not be compared across benchmarks; in particular, FollowBench rule mean should be interpreted according to its scoring definition.

Figure 4: Upper-layer geometry diagnostics. Top modules by geometric similarity between instruction- and numeric-reasoning-tuned LoRA updates. This figure provides preliminary evidence but does not establish a localized causal mechanism.

Figure 5: Functional probing summary with null or small effects. Largest observed effect sizes from localized functional probing. We do not find strong localized causal evidence that explains the mismatch pattern in this setup.

Stage	Model	Setting	$n$	Drift score (mean $\pm$ std)
benchmark_suite	llama31_8b_instruct	r16_attnmlp_do005_lrmain	1	$0.485$
benchmark_suite	qwen3_14b	r16_attnmlp_do005_lrmain	1	$0.674$
benchmark_suite	qwen3_8b	r16_attnmlp_do005_lrmain	1	$0.606$
model_sweep	deepseek_r1_distill_qwen_7b	r16_attnonly_do0_lrsmall	1	$0.284$
model_sweep	deepseek_r1_distill_qwen_7b	r32_attnmlp_do005_lrmain	1	$0.350$
model_sweep	llama31_8b_instruct	r16_attnmlp_do005_lrmain	2	$0.394\pm 0.016$

Appendix

Table 6: Robustness: drift score summary (instr

\rightarrow

numeric reasoning). Drift score mean and standard deviation across seeds (when

n>1

); values are rounded to three decimals.

Instruction Tuning

Instruction tuning adapts pretrained language models to follow natural-language requests using curated or synthesized instruction-response examples.

See all 82 papers

Apr 7, 2026

How LLMs Follow Instructions: Skillful Coordination, Not a Universal Mechanism

Instruction tuning is commonly assumed to endow language models with a domain-general ability to follow instructions, yet the underlying mechanism remains poorly understood. Does instruction-following rely on a universal mechanism or compositional skill deployment? We investigate this through diagnostic probing across nine diverse tasks in three instruction-tuned models. Our analysis provides converging evidence against a universal mechanism. First, general probes trained across all tasks show selective rather than uniform deficits relative to task-specific specialists, indicating that representational sharing is partial and structured rather than global. Second, cross-task transfer is weak and clustered by skill similarity. Third, causal ablation reveals sparse asymmetric dependencies rather than shared representations. Tasks also stratify by complexity across layers, with structural constraints emerging early and semantic tasks emerging late. Finally, temporal analysis shows that the constraint signal becomes decodable only once generation is under way, and remains so throughout the response. These findings indicate that instruction-following is better characterized as skillful coordination of diverse linguistic capabilities rather than deployment of a single abstract constraint-checking process.

Elisabetta Rocchetti, Alfio Ferrara

Department of Computer Science, Università degli Studi di Milano

LLM Interpretability Instruction Following Instruction Tuning

Ask PDF

Jun 9, 2026

Which LoRA? An Empirical Study on the Effectiveness of LoRA Techniques During Multilingual Instruction Tuning

We investigate whether commonly available LoRA variants have an advantage over basic LoRA in multilingual instruction tuning. Experiments involving LoRA and four other variants on two datasets across diverse target languages show that there is no significant advantage in using more complex LoRA variants instead of basic LoRA, with respect to balancing cross-lingual transfer and knowledge retention. An analysis of hidden embeddings reveal that layer-wise language representation remains largely similar across LLMs fine-tuned with different LoRA techniques, suggesting that architectural novelty of LoRA techniques may not translate into better cross-lingual adaptation.

Thamali Wijewardhana, Napoleon H. Reyes, Surangika Ranathunga

School of Mathematical and Computational Sciences Massey University, Auckland, New Zealand

Low-Rank Adaptation Cross-Lingual Transfer

Ask PDF

Apr 10, 2026

Automated Instruction Revision (AIR): A Structured Comparison of Task Adaptation Strategies for LLM

This paper studies Automated Instruction Revision (AIR), a rule-induction-based method for adapting large language models (LLMs) to downstream tasks using limited task-specific examples. We position AIR within the broader landscape of adaptation strategies, including prompt optimization, retrieval-based methods, and fine-tuning. We then compare these approaches across a diverse benchmark suite designed to stress different task requirements, such as knowledge injection, structured extraction, label remapping, and logical reasoning. The paper argues that adaptation performance is strongly task-dependent: no single method dominates across all settings. Across five benchmarks, AIR was strongest or near-best on label-remapping classification, while KNN retrieval performed best on closed-book QA, and fine-tuning dominated structured extraction and event-order reasoning. AIR is most promising when task behavior can be captured by compact, interpretable instruction rules, while retrieval and fine-tuning remain stronger in tasks dominated by source-specific knowledge or dataset-specific annotation regularities.

Solomiia Bilyk, Volodymyr Getmanskyi, Taras Firman

Eleks Ltd. · Ukrainian Catholic University, Lviv, Ukraine

LLM Fine-Tuning Prompt Optimization

Ask PDF

Abstract

Instruction Tuning

Explore similar work

How LLMs Follow Instructions: Skillful Coordination, Not a Universal Mechanism

Which LoRA? An Empirical Study on the Effectiveness of LoRA Techniques During Multilingual Instruction Tuning

Automated Instruction Revision (AIR): A Structured Comparison of Task Adaptation Strategies for LLM