iGEN
Visit IGEN World Explore IGEN Expo
EXPLORE UPGRADE PLANS
BREAKING
Werner Enterprises Posts Highest Revenue Per Truck Growth in One-Way Segment in a Decade CMA CGM and Stonepeak Launch United Ports LLC in $2.4 Billion Terminal Joint Venture UPS shift away from Amazon shows bigger payoff Lanesurf: 62% of Loads Get Vetted Carrier Offers Before Brokers Arrive India-China Border Trade Via Lipulekh Resumes Aug 1; China Permits 20 Traders Geopolitics Drives CMA CGM Q2 Profit Surge of 42% as Volumes and Rates Climb Benchmark Diesel Price Rises Third Week as Futures Plunge; Spread Hits Record Indian Government Limits Sugar Dealers to 400 Tonnes Stock Until November to Curb Hoarding Tenants signing longer leases for larger warehouses as 3PLs lock in capacity US stock market flat as S&P 500 and Dow barely move, Nasdaq slides over 1% on chip rout Werner Enterprises Posts Highest Revenue Per Truck Growth in One-Way Segment in a Decade CMA CGM and Stonepeak Launch United Ports LLC in $2.4 Billion Terminal Joint Venture UPS shift away from Amazon shows bigger payoff Lanesurf: 62% of Loads Get Vetted Carrier Offers Before Brokers Arrive India-China Border Trade Via Lipulekh Resumes Aug 1; China Permits 20 Traders Geopolitics Drives CMA CGM Q2 Profit Surge of 42% as Volumes and Rates Climb Benchmark Diesel Price Rises Third Week as Futures Plunge; Spread Hits Record Indian Government Limits Sugar Dealers to 400 Tonnes Stock Until November to Curb Hoarding Tenants signing longer leases for larger warehouses as 3PLs lock in capacity US stock market flat as S&P 500 and Dow barely move, Nasdaq slides over 1% on chip rout
Home ›› Technology ›› Ai ›› Llms ›› Sequential DPO Study Reveals Non-Uniform Forgetting Across Multiple Preference Objectives

Sequential DPO Study Reveals Non-Uniform Forgetting Across Multiple Preference Objectives

A study by Bhandari et al. on sequential Direct Preference Optimization (DPO) finds that later training objectives do not uniformly degrade earlier preferences. Using Llama-3.1-8B-Instruct, the research reveals that forgetting patterns vary from stability to positive transfer depending on objective compatibility and signal strength, offering guidance for multi-objective AI alignment in enterprises.

iG
iGEN Editorial
July 8, 2026
Sequential DPO Study Reveals Non-Uniform Forgetting Across Multiple Preference Objectives

Enterprises deploying large language models (LLMs) often need to align them with multiple, sometimes conflicting human preferences — safety, helpfulness, factual accuracy, and brand voice. A common approach is to apply these preferences sequentially using methods like Direct Preference Optimization (DPO). But does later training erase what the model learned earlier? According to a study by Bhandari, Pranav; Fay, Nicolas; Datta, Amitava; Naseem, Usman; and Nasim, Mehwish on arXiv, the answer is not straightforward: forgetting is not uniform.

Four Preference Settings

The researchers studied sequential DPO across four distinct preference settings:

Setting Description
Distributional conflict Objectives that compete due to differing data distributions
Multi-attribute interaction Preferences involving multiple attributes that interact
Strong safety signal A high-priority safety objective
Compatible response-quality Objectives that are naturally aligned

Using the Llama-3.1-8B-Instruct model with LoRA adapters, the team evaluated all objectives after every stage while keeping a fixed base-model reference.

Key Findings on Forgetting

The central finding is that sequential DPO does not produce a single forgetting pattern. According to the study, preference change ranges from "partial degradation to stability, pair-level redistribution, or positive transfer" depending on the relationship between objectives, signal strength, and training order. This means that aggregate metrics can mask heterogeneous changes across individual preference pairs.

To uncover these patterns, the authors used pair-level analysis with length-normalized policy margins. They also applied quartile decomposition, which showed that high-confidence pairs could either degrade or improve depending on the setting. Mechanistic diagnostics revealed that Stage 2 gradients and adapter updates are near-orthogonal to the previous objective across all settings, providing little evidence that direct gradient opposition drives forgetting.

sequential DPO does not produce a single forgetting pattern; preference change ranges from partial degradation to stability, pair-level redistribution, or positive transfer

Implications for Enterprise AI

For enterprise technology leaders, these findings underscore that sequential alignment pipelines must account for objective compatibility and signal strength. Assuming later objectives uniformly overwrite earlier preferences is incorrect. Instead, careful design of the sequence and monitoring of pair-level performance are needed to avoid unintended degradation of critical preferences, such as safety or compliance.

The study recommends that future alignment pipelines should evaluate objective relationships and training order rather than treating all later training as uniformly harmful. This is particularly relevant for organizations that deploy LLMs in regulated industries where multiple behavioral constraints must coexist.


Sources:

Keep Reading

Recommended Stories

New Research Reveals How Visual Tokens Evolve Inside Vision-Language Models Technology

New Research Reveals How Visual Tokens Evolve Inside Vision-Language Models

A new computer vision paper from arXiv investigates how visual tokens are integrated into large language models (LLMs) under two paradigms: in-context prompting and layer-wise injection. The authors find that visual tokens enter the LLM as 'disguised visual context' lacking linguistic structure, then evolve differently depending on the integration architecture. They show that attention allocation alone is insufficient, and performance depends on the quality of visual representations at each layer.

July 8, 2026
Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models Technology

Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models

A new research paper from arXiv shows that reinforcement learning with verifiable rewards (RLVR) can cause large reasoning models to forget foundational capabilities like perception and faithfulness. The authors propose RECAP, a replay strategy with dynamic objective reweighting that preserves general knowledge while maintaining reasoning gains.

June 21, 2026
SafeSpec: New Framework Boosts LLM Safety Without Sacrificing Inference Speed Technology

SafeSpec: New Framework Boosts LLM Safety Without Sacrificing Inference Speed

Researchers propose SafeSpec, a safety-aware speculative inference framework that attaches a latent safety head to jointly evaluate semantic validity and safety in a single forward pass. On Qwen3-32B, it reduces attack success rates by 15% while preserving a 2.06x inference speedup on benign workloads, addressing the fundamental incompatibility between existing safety methods and speculative decoding.

June 21, 2026
ArXiv Paper Introduces KG-SoftMAP: Bayesian Network Learning from Sparse Data Using Knowledge Graph Priors Technology

ArXiv Paper Introduces KG-SoftMAP: Bayesian Network Learning from Sparse Data Using Knowledge Graph Priors

KG-SoftMAP is a new method for Bayesian network structure learning from sparse discrete data, using a weighted knowledge graph as a prior. On synthetic benchmarks, it achieved Directed-F1 scores up to 0.97 at higher observation rates, while data-only learners stayed near zero. On real educational datasets, it matched logistic regression within 0.03 F1_FAIL while providing an interpretable concept graph.

June 21, 2026