When the Model Said 'No Comment', We Knew Helpfulness Was Dead, Honesty Was Alive, and Safety Was Terrified
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911430744211456 |
|---|---|
| author | Kashyap, Gautam Siddharth Dras, Mark Naseem, Usman |
| author_facet | Kashyap, Gautam Siddharth Dras, Mark Naseem, Usman |
| contents | Large Language Models (LLMs) need to be in accordance with human values-being helpful, harmless, and honest (HHH)-is important for safe deployment. Existing works use Supervised Fine-Tuning (SFT) and Mixture-of-Experts (MoE) to align LLMs. However, these works face challenges in multi-objective settings, such as SFT leading to interference between conflicting objectives, while MoEs suffer from miscalibrated routing. We term this failure mode Axis Collapse, marked by (1) disjoint feature spaces causing catastrophic forgetting, and (2) unreliable inference from misrouted experts. To resolve this, we propose AlignX, a two-stage framework. Stage 1 uses prompt-injected fine-tuning to extract axis-specific task features, mitigating catastrophic forgetting. Stage 2 deploys a MoCaE module that calibrates expert routing using fractal and natural geometry, improving inference reliability. AlignX achieves significant gains on Alpaca (Helpfulness), BeaverTails (Harmlessness), and TruthfulQA (Honesty), with +171.5% win rate, +110.1% in truthfulness-informativeness, and 4.3% fewer safety violations. It also reduces latency and memory usage by over 35% compared to prior MoEs. Results across four LLMs validate its generalizability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_07381 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | When the Model Said 'No Comment', We Knew Helpfulness Was Dead, Honesty Was Alive, and Safety Was Terrified Kashyap, Gautam Siddharth Dras, Mark Naseem, Usman Computation and Language Large Language Models (LLMs) need to be in accordance with human values-being helpful, harmless, and honest (HHH)-is important for safe deployment. Existing works use Supervised Fine-Tuning (SFT) and Mixture-of-Experts (MoE) to align LLMs. However, these works face challenges in multi-objective settings, such as SFT leading to interference between conflicting objectives, while MoEs suffer from miscalibrated routing. We term this failure mode Axis Collapse, marked by (1) disjoint feature spaces causing catastrophic forgetting, and (2) unreliable inference from misrouted experts. To resolve this, we propose AlignX, a two-stage framework. Stage 1 uses prompt-injected fine-tuning to extract axis-specific task features, mitigating catastrophic forgetting. Stage 2 deploys a MoCaE module that calibrates expert routing using fractal and natural geometry, improving inference reliability. AlignX achieves significant gains on Alpaca (Helpfulness), BeaverTails (Harmlessness), and TruthfulQA (Honesty), with +171.5% win rate, +110.1% in truthfulness-informativeness, and 4.3% fewer safety violations. It also reduces latency and memory usage by over 35% compared to prior MoEs. Results across four LLMs validate its generalizability. |
| title | When the Model Said 'No Comment', We Knew Helpfulness Was Dead, Honesty Was Alive, and Safety Was Terrified |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2602.07381 |